DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Receipt is acknowledged of certified copies of papers submitted under 35 U.S.C. 119(a)-(d), which papers have been placed of record in the file.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 09/13/2024 and 03/04/2025 have been considered by the examiner.
Claim Objections
Claim 18 is objected to because of the following informalities:
In claim 18, line 14, “the processor makes sets a signal value…” should read “the processor
Appropriate correction is required.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
Claims 1-7, 9-11, and 13 recite limitations that use words like “means” (or “step”) or similar terms with functional language and do invoke 35 U.S.C. 112(f):
Claim 1; recites the limitation, “a first step of generating… using a machine learning model” [Lines 2-3].
Claim 1; recites the limitation, “a second step of generating a third image…” [Lines 4-5].
Claim 2; recites the limitation, “wherein the second step sets the signal value of each pixel…” [Lines 1-2].
Claims 3, 4, 5, 7, 13; recite the limitation, “wherein in the second step, the third image is generated…” [Lines 1-2].
Claim 6; recites the limitation, “wherein in the second step, the second residual image is generated…” [Lines 1-2].
Claim 9; recites the limitation, “wherein in the first step, the number of pixels…” [Lines 1-2].
Claims 10, 11; recite the limitation, “wherein in the first step, the input data input…” [Lines 1-2].
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
After a careful analysis, as disclosed above, and a careful review of the specification the following limitations in claims 1-7, 9-11, and 13:
“first step” (Fig. 1, Paragraph [0044], – “Next, the upsampling using a trained CNN (machine learning model) executed in the image pickup apparatus 102 will be described with reference to FIG. 1 illustrating the flow of the processing and a flowchart in FIG. 6. In the image pickup apparatus 102, the acquiring unit 125 and the calculator 126 constituted by a computer (image processing apparatus) execute the following processing (first and second steps) according to a (computer) program.” See also Figs. 5A-B, Paragraphs [0047-0050]. Thus, the first step has sufficient structure wherein it is carried out by a computer.
“second step” (Fig. 1, Paragraph [0044], – “Next, the upsampling using a trained CNN (machine learning model) executed in the image pickup apparatus 102 will be described with reference to FIG. 1 illustrating the flow of the processing and a flowchart in FIG. 6. In the image pickup apparatus 102, the acquiring unit 125 and the calculator 126 constituted by a computer (image processing apparatus) execute the following processing (first and second steps) according to a (computer) program.” See also Figs. 5A-B, Paragraphs [0047-0050]. Thus, the second step has sufficient structure wherein it is carried out by a computer.
“machine learning model” (Figs. 5A-B, Paragraph [0036] – “Example 1 uses the Convolutional Neural Network (CNN) illustrated in FIGs. 5A and 5B as the machine learning model. However, the type of machine learning model is not limited to this example, and Multi-Layer Perceptron (MLP), Vision Transformer (ViT), Generative Adversarial Network (GAN), diffusion model, etc. may be used.” Thus, the machine learning model has sufficient structure wherein it is any one of a convolutional neural network, multi-layer perceptron, vision transformer, generative adversarial network, diffusion model, or the like.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed
invention is not identically disclosed as set forth in section 102 of this title, if the
differences between the claimed invention and the prior art are such that the claimed
invention as a whole would have been obvious before the effective filing date of the
claimed invention to a person having ordinary skill in the art to which the claimed
invention pertains. Patentability shall not be negated by the manner in which the
invention was made.
Claims 1, 2, 4, 9, 13, 14, 16, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS.
Regarding claim 1, SU teaches an image processing method (Fig. 7, Paragraph [0006] – SU discloses there is provided a computer-implemented method for increasing image resolution of a digital image.) comprising:
a first step of generating a first residual image based on a first image (Fig. 3, Paragraph [0076] – SU discloses the NN model 306 uses multiple convolutional layers to involve the LR residual image 304 and generate a plurality of HR residual sub-images 308 corresponding to the input LR image 302 [wherein input LR image 302 is a first image]. In some aspects, the plurality of HR residual sub-images 308 includes four sub-images. The pixel shifting module 318 performs pixel shifting on the plurality of HR residual sub-images 308 to generate an HR residual image 310 [wherein HR residual image 310 is a first residual image].)
using a machine learning model (Fig. 3, #306 called NN model, Paragraph [0017] – SU discloses the neural network model includes an input layer and the plurality of convolutional layers comprises four convolutional layers. The input layer is configured to receive the digital image. An output layer of the four convolutional layers is configured to output the plurality of HR residual sub-images. Paragraph [0078] – SU further discloses one or more functionalities performed by the bicubic upsampling module 316, the residue generation module 322, the pixel shifting module 318, and the adder 320 can be performed by the NN model 306.);
wherein the fourth image is the first image or an image based on the first image (Fig. 3, Paragraph [0076] – SU discloses a low-resolution (LR) input image 302 [wherein input LR image 302 is the first image] is processed (e.g., by the bicubic upsampling module 316 and the residue generation module 322) to generate a base HR image 312 [wherein base HR image 312 is the fourth image].),
Although SU further teaches and wherein in a case where a second image is a sum of the fourth image and the first residual image (Fig. 3, Paragraph [0076] – SU discloses adder 320 may comprise suitable circuitry, logic, interfaces, or code and is configured to add the base HR image 312 [wherein base HR image 312 is the fourth image] with the HR residual image 310 [wherein HR residual image 310 is the first residual image] to generate an HR image 314 [wherein HR image 314 is a second image] as an output image corresponding to the input LR image 302.),
SU fails to explicitly teach and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, the second step sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
However, SHROERS explicitly teaches and a second step of generating a third image (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image] using a respective one of inpainted masked patches 354a-354l corresponding to the location of the one or more anomalous pixels in input image 230/330/530 that were identified in action 465 (action 468).)
based on a fourth image (Fig. 4, Paragraph [0049] - SHROERS discloses flowchart 460 continues with identifying one or more anomalous pixels in input image 230/330/530 [wherein input image is a fourth image] using residual image 236/536 (action 465) [wherein residual image is the first residual image].)
and an absolute value of the first residual image (Fig. 4, Paragraph [0039] – SHROERS further discloses pixels in residual image 236 may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330.),
the second step sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image (Fig. 4, Paragraph [0042] – SHROERS discloses action 465 may include detecting one or more anomaly candidates in input image 230/330/530 using residual image 236/536, based, for example, on the brightness of the one or more anomaly candidates, and determining a residual value associated with each of the one or more anomaly candidates, such as a residual value corresponding to its relative brightness. Paragraph [0039] – SHROERS further discloses where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch.)
as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases (Fig. 2, Paragraph [0039] – SHROERS discloses pixels in residual image 236 [wherein residual image 236 is a first residual image] may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330. Thus, where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch. [See also Fig. 4.]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU of having an image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image, with the teachings of SHROERS having and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, the second step sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
Wherein SU’s image processing method wherein having and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, the second step sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and automate pixel error detection, since both SU and SHROERS relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and SHROERS relates to systems and methods for performing automated pixel error detection using an inpainting neural network, the present solution advantageously yields better results than other attempts to automate pixel error detection, with lower false positive rates and very high recall, and in fact enables the detection of pixel errors that human inspectors and previous automated pixel error detection solutions had failed to identify. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and SHROERS (US 20210304387 A1), Paragraph [0003, 0056].
Regarding claim 2, SU in view of SHROERS teach the image processing method according to claim 1,
SU fails to explicitly teach wherein the second step sets the signal value of each pixel in the third image to a value between the signal value of the pixel in the fourth image corresponding to each pixel in the third image and the signal value of the pixel in the second image corresponding to each pixel in the third image.
However, SHROERS explicitly teaches wherein the second step sets the signal value of each pixel in the third image (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image] using a respective one of inpainted masked patches 354a-354l corresponding to the location of the one or more anomalous pixels in input image 230/330/530 that were identified in action 465 (action 468). See also Paragraph [0053].)
to a value between the signal value of the pixel in the fourth image corresponding to each pixel in the third image and the signal value of the pixel in the second image corresponding to each pixel in the third image (Fig. 4, Paragraph [0042] – SHROERS discloses action 465 may include detecting one or more anomaly candidates in input image 230/330/530 using residual image 236/536, based, for example, on the brightness of the one or more anomaly candidates, and determining a residual value associated with each of the one or more anomaly candidates, such as a residual value corresponding to its relative brightness [wherein a residual value corresponding to relative brightness is a signal value]. Paragraph [0039] – SHROERS further discloses where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch. See also Paragraph [0052].).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of SHROERS of having wherein the second step sets the signal value of each pixel in the third image to a value between the signal value of the pixel in the fourth image corresponding to each pixel in the third image and the signal value of the pixel in the second image corresponding to each pixel in the third image.
Wherein SU’s image processing method wherein the second step sets the signal value of each pixel in the third image to a value between the signal value of the pixel in the fourth image corresponding to each pixel in the third image and the signal value of the pixel in the second image corresponding to each pixel in the third image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and automate pixel error detection, since both SU and SHROERS relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and SHROERS relates to systems and methods for performing automated pixel error detection using an inpainting neural network, the present solution advantageously yields better results than other attempts to automate pixel error detection, with lower false positive rates and very high recall, and in fact enables the detection of pixel errors that human inspectors and previous automated pixel error detection solutions had failed to identify. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and SHROERS (US 20210304387 A1), Paragraph [0003, 0056].
Regarding claim 4, SU in view of SHROERS teach the image processing method according to claim 1,
SU fails to explicitly teach wherein in the second step, the third image is generated using a result of threshold processing based on the absolute value of the first residual image.
However, SHROERS explicitly teaches wherein in the second step, the third image is generated (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image] using a respective one of inpainted masked patches 354a-354l corresponding to the location of the one or more anomalous pixels in input image 230/330/530 that were identified in action 465 (action 468).)
using a result of threshold processing (Fig. 4, Paragraph [0042] – SHROERS discloses the one or more anomaly candidates may be identified as actually anomalous pixels based on comparing their residual values with a predetermined threshold residual value. For example, where the residual values correspond to brightness, the brightness of each anomaly candidate may be compared to a predetermined brightness threshold, mid only those anomaly candidates meeting or exceeding the brightness threshold are identified as anomalous in action 465.).)
based on the absolute value of the first residual image (Fig. 4, Paragraph [0039] – SHROERS further discloses pixels in residual image 236 may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of SHROERS of having wherein in the second step, the third image is generated using a result of threshold processing based on the absolute value of the first residual image.
Wherein SU’s image processing method wherein in the second step, the third image is generated using a result of threshold processing based on the absolute value of the first residual image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and automate pixel error detection, since both SU and SHROERS relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and SHROERS relates to systems and methods for performing automated pixel error detection using an inpainting neural network, the present solution advantageously yields better results than other attempts to automate pixel error detection, with lower false positive rates and very high recall, and in fact enables the detection of pixel errors that human inspectors and previous automated pixel error detection solutions had failed to identify. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and SHROERS (US 20210304387 A1), Paragraph [0003, 0056].
Regarding claim 9, SU in view of SHROERS teach the image processing method according to claim 1,
SU further teaches wherein in the first step (Fig. 3, Paragraph [0076] – SU discloses the NN model 306 uses multiple convolutional layers to involve the LR residual image 304 and generate a plurality of HR residual sub-images 308 corresponding to the input LR image 302. The pixel shifting module 318 performs pixel shifting on the plurality of HR residual sub-images 308 to generate an HR residual image 310),
the number of pixels in a vertical or horizontal direction in any of input data input to the machine learning model (Fig. 3, Paragraph [0074] – SU discloses NN model 306 is configured to generate a plurality of HR residual sub-images 308 corresponding to an input LR image 302 based on an LR residual image (e.g., a grayscale version of the LR input image such as LR residual image 304). Paragraph [0042] – SU further discloses the terms “low-resolution” (or LR) and “high-resolution” (or HR) in connection with an image are associated with the size of the image (in pixels). For example, if two images depict the same scene and the first image has bigger height and width (in pixels) in comparison to the height and width of the second image, then the first image is referred to as a high-resolution image and the second image is referred to as a low-resolution image. Alternatively, a high-resolution image has a high pixel density than an LR image, such as where the HR image has a large number of pixels per inch (or another distance measurement).),
output data output from the machine learning model (Fig. 3, Paragraph [0074] – SU discloses NN model 306 is configured to generate a plurality of HR residual sub-images 308 corresponding to an input LR image 302 based on an LR residual image (e.g., a grayscale version of the LR input image such as LR residual image 304). Paragraph [0042] – SU further discloses the terms “low-resolution” (or LR) and “high-resolution” (or HR) in connection with an image are associated with the size of the image (in pixels). For example, if two images depict the same scene and the first image has bigger height and width (in pixels) in comparison to the height and width of the second image, then the first image is referred to as a high-resolution image and the second image is referred to as a low-resolution image. Alternatively, a high-resolution image has a high pixel density than an LR image, such as where the HR image has a large number of pixels per inch (or another distance measurement).),
and a plurality of feature maps generated within the machine learning model (Fig. 3, Paragraph [0076] – SU discloses NN model 306 uses multiple convolutional layers to involve the LR residual image 304 and generate a plurality of HR residual sub-images 308 corresponding to the input LR image 302.)
is different from another (Fig. 3, Paragraph [0042] – SU further discloses the terms “low-resolution” (or LR) and “high-resolution” (or HR) in connection with an image are associated with the size of the image (in pixels). For example, if two images depict the same scene and the first image has bigger height and width (in pixels) in comparison to the height and width of the second image, then the first image is referred to as a high-resolution image and the second image is referred to as a low-resolution image. Alternatively, a high-resolution image has a high pixel density than an LR image, such as where the HR image has a large number of pixels per inch (or another distance measurement).).
Regarding claim 13, SU in view of SHROERS teach the image processing method according to claim 1,
SU fails to explicitly teach wherein in the second step, the third image is generated based on information about an edge of the fourth image or a magnitude of a signal value of each pixel in the fourth image.
However, SHROERS explicitly teaches wherein in the second step, the third image is generated (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image].)
based on information about an edge of the fourth image (Fig. 4, Paragraph [0038] – SHROERS discloses although flowchart 460 is directed to an automated method for performing pixel error detection, in some implementations it may be advantageous or desirable to enable a system user to select one or more patches of an input image for analysis. For example, in some use cases regions at the edges of an input image [wherein input image is the fourth image] may be considered less (or more) important than regions closer to the center of the input image.)
or a magnitude of a signal value of each pixel in the fourth image (Fig. 4, Paragraph [0039] – SHROERS discloses pixels in residual image 236 may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of SHROERS of having wherein in the second step, the third image is generated based on information about an edge of the fourth image or a magnitude of a signal value of each pixel in the fourth image.
Wherein SU’s image processing method wherein in the second step, the third image is generated based on information about an edge of the fourth image or a magnitude of a signal value of each pixel in the fourth image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and restoration details, since both SU and ZHANG relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and ZHANG relates to a single image dehazing method based on detail recovery; while effectively removing fog, it can retain more image detail information, reduce the blurring at the edges of the defogged image, and generate higher-quality defogged images. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and ZHANG (US 20240289928 A1), Paragraph [0005].
Regarding claim 14, SU in view of SHROERS teach the image processing method according to claim 1,
wherein the second image is an image obtained by performing at least one of degradation correcting (Fig. 3, Paragraph [0092] – SU discloses using the neural network model training techniques discussed herein (e.g., degrading the image quality on the input side and enhancing the image quality on the output side), the NN model 306 can generate images or video frames of superior visual perception during UR processing of LR images or video frames.),
SU further teaches resolution improving (Fig. 3, Paragraph [0043] – SU discloses “super-resolution” (or SR) refers to a resolution enhancement technique that increases the number of pixels (e.g., via upscaling) and improves the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) compared to conventional interpolation methods. Paragraph [0092] – SU further discloses the NN model 306 can generate images or video frames of superior visual perception during UR processing of LR images or video frames. Additionally, the use of a concise neural network model (e.g., convolutional layers with a limited number of layers, such as four, within the NN model 306) that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices (or other types of limited-resource devices, such as smart TVs, tablets, laptops, and other computing devices) for performing real-time UR processing.),
contrast improving (Fig. 3, Paragraph [0092] – SU discloses the NN model 306 can generate images or video frames of superior visual perception during UR processing of LR images or video frames. Additionally, the use of a concise neural network model (e.g., convolutional layers with a limited number of layers, such as four, within the NN model 306) that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices (or other types of limited-resource devices, such as smart TVs, tablets, laptops, and other computing devices) for performing real-time UR processing.),
demosaicing, dynamic range expanding, gradation increasing, lighting changing, changing a defocus blur shape, focus distance changing, and depth of field changing for the first image (Fig. 3, Paragraph [0092] – SU discloses using the neural network model training techniques discussed herein (e.g., degrading the image quality on the input side and enhancing the image quality on the output side), the NN model 306 can generate images or video frames of superior visual perception during UR processing of LR images or video frames.).
Regarding claim 16, SU in view of SHROERS teach the image processing method according to claim 1,
SU further teaches a non-transitory computer-readable storage medium storing a program for causing a computer to execute the image processing method (Fig. 9, Paragraph [0112] – SU discloses computer-readable instructions stored on a computer-readable medium (e.g., the program 955 stored in the memory 910) are executable by the processor 905 of the computing device 900. A hard drive, CD-ROM, and RAM are some examples of articles including a non-transitory computer-readable medium such as a storage device. “Computer-readable non-transitory media” includes all types of computer-readable media, including magnetic storage media, optical storage media, flash media, and solid-state storage media. It should be understood that software can be installed in and sold with a computer. See also Paragraph [0109].) according to claim 1.
Regarding claim 17, SU teaches an image processing apparatus (Fig. 9, #900 called computer, Paragraph [0109-0110] – SU discloses FIG. 9 is a block diagram illustrating circuitry for a device that implements algorithms and performs methods, according to example embodiments. One example computing device in the form of a computer 900.) comprising:
a memory storing instructions (Fig. 9, #910 called memory, Paragraph [0110] – SU discloses computer 900 (also referred to as computing device 900, computer system 900, or computer 900) may include a processor 905, memory 910. Paragraph [0111] - SU further discloses memory 910 may include volatile memory 945 and non-volatile memory 950 and may store a program 955.);
a processor (Fig. 9, #905 called processor, Paragraph [0110]) for executing the instructions (Fig. 9, Paragraph [0110] – SU discloses computer 900 (also referred to as computing device 900, computer system 900, or computer 900) may include a processor 905. Paragraph [0112] – SU further discloses computer-readable instructions stored on a computer-readable medium (e.g., the program 955 stored in the memory 910) are executable by the processor 905 of the computing device 900.) to:
generate a first residual image based on a first image (Fig. 3, Paragraph [0076] – SU discloses the NN model 306 uses multiple convolutional layers to involve the LR residual image 304 and generate a plurality of HR residual sub-images 308 corresponding to the input LR image 302 [wherein input LR image 302 is a first image]. In some aspects, the plurality of HR residual sub-images 308 includes four sub-images. The pixel shifting module 318 performs pixel shifting on the plurality of HR residual sub-images 308 to generate an HR residual image 310 [wherein HR residual image 310 is a first residual image].)
using a machine learning model (Fig. 3, #306 called NN model, Paragraph [0017] – SU discloses the neural network model includes an input layer and the plurality of convolutional layers comprises four convolutional layers. The input layer is configured to receive the digital image. An output layer of the four convolutional layers is configured to output the plurality of HR residual sub-images. Paragraph [0078] – SU further discloses one or more functionalities performed by the bicubic upsampling module 316, the residue generation module 322, the pixel shifting module 318, and the adder 320 can be performed by the NN model 306.);
wherein the fourth image is the first image or an image based on the first image (Fig. 3, Paragraph [0076] – SU discloses a low-resolution (LR) input image 302 [wherein input LR image 302 is the first image] is processed (e.g., by the bicubic upsampling module 316 and the residue generation module 322) to generate a base HR image 312 [wherein base HR image 312 is the fourth image].),
Although SU further teaches and wherein in a case where a second image is a sum of the fourth image and the first residual image (Fig. 3, Paragraph [0076] – SU discloses adder 320 may comprise suitable circuitry, logic, interfaces, or code and is configured to add the base HR image 312 [wherein base HR image 312 is the fourth image] with the HR residual image 310 [wherein HR residual image 310 is the first residual image] to generate an HR image 314 [wherein HR image 314 is a second image] as an output image corresponding to the input LR image 302.),
SU fails to explicitly teach and generate a third image based on a fourth image and an absolute value of the first residual image, the processor sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
However, SHROERS explicitly teaches and generate a third image (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image] using a respective one of inpainted masked patches 354a-354l corresponding to the location of the one or more anomalous pixels in input image 230/330/530 that were identified in action 465 (action 468).)
based on a fourth image (Fig. 4, Paragraph [0049] - SHROERS discloses flowchart 460 continues with identifying one or more anomalous pixels in input image 230/330/530 [wherein input image is a fourth image] using residual image 236/536 (action 465) [wherein residual image is the first residual image].)
and an absolute value of the first residual image (Fig. 4, Paragraph [0039] – SHROERS further discloses pixels in residual image 236 may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330.),
the processor (Fig. 1, #104 called hardware processor, Paragraph [0014]) sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image (Fig. 4, Paragraph [0042] – SHROERS discloses identification of those bright pixels as one or more anomalous pixels may be performed by software code 110, executed by hardware processor 104. Action 465 may include detecting one or more anomaly candidates in input image 230/330/530 using residual image 236/536, based, for example, on the brightness of the one or more anomaly candidates, and determining a residual value associated with each of the one or more anomaly candidates, such as a residual value corresponding to its relative brightness. Paragraph [0039] – SHROERS further discloses where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch.)
as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases (Fig. 2, Paragraph [0039] – SHROERS discloses pixels in residual image 236 [wherein residual image 236 is a first residual image] may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330. Thus, where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch. [See also Fig. 4.]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU of having an image processing apparatus comprising: a memory storing instructions; a processor for executing the instructions to: generate a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image, with the teachings of SHROERS having and generate a third image based on a fourth image and an absolute value of the first residual image, the processor sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
Wherein SU’s image processing apparatus wherein having and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, the second step sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
The motivation behind this modification would have been to provide an enhanced image processing apparatus that can improve image resolution and automate pixel error detection, since both SU and SHROERS relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and SHROERS relates to systems and methods for performing automated pixel error detection using an inpainting neural network, the present solution advantageously yields better results than other attempts to automate pixel error detection, with lower false positive rates and very high recall, and in fact enables the detection of pixel errors that human inspectors and previous automated pixel error detection solutions had failed to identify. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and SHROERS (US 20210304387 A1), Paragraph [0003, 0056].
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS in further view of SU, G (US 20140050271 A1), hereinafter referenced as SU, G.
Regarding claim 3, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein in the second step, the third image is generated using a result of a nonlinear transformation based on the absolute value of the first residual image.
However, SU, G explicitly teaches wherein in the second step, the third image is generated (Fig. 3, Paragraph [0038] – SU, G discloses residual 332 is decoded (340), de-quantized (350), and added to the output 395 of the predictor 390 to generate the output VDR signal 370 [wherein the output VDR signal is a third image]. See also Paragraph [0031].)
using a result of a nonlinear transformation (Fig. 5A-5B, Paragraph [0036] – SU, G discloses a novel non-linear quantizer based on the characteristics of sigmoid transfer functions, such as the mu-law (μ-law) transfer function, is described. Paragraph [0038] – SU, G further discloses non-linear de-quantizers based on the characteristics of sigmoid transfer functions, such as the mu-law (μ-law) transfer function, are described.).
based on the absolute value of the first residual image (Fig. 1, Paragraph [0044-0045] – SU, G discloses the μ-law and A-law forward and inverse transfer functions are described as: Mu-Law
PNG
media_image1.png
269
836
media_image1.png
Greyscale
where xmax denotes the maximum absolute value of the input signal. See also paragraph [0047].).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of SU, G having wherein in the second step, the third image is generated using a result of a nonlinear transformation based on the absolute value of the first residual image.
Wherein SU’s image processing method wherein in the second step, the third image is generated using a result of a nonlinear transformation based on the absolute value of the first residual image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and overall video quality, since both SU and SU, G relate to image and video processing techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and SU, G relates to relates to the non-linear quantization and de-quantization of the residual signal in layered coding of high dynamic range images, thus yielding advantageous coding efficiency and improved overall video quality. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and SU, G (US 20140050271 A1), Paragraph [0002, 0026].
Claims 5-8 are rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS in further view of ZHANG (US 20240289928 A1), hereinafter referenced as ZHANG.
Regarding claim 5, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein in the second step, the third image is generated based on a second residual image generated from the first residual image using soft thresholding.
However, ZHANG explicitly teaches wherein in the second step, the third image is generated based on a second residual image generated from the first residual image using soft thresholding (Fig. 5, Paragraph [0043] – ZHANG discloses the residual shrinkage module includes two 3×3 convolutional layers, a global mean pooling layer, and a fully connected layer, inputting feature images into the two 3×3 convolutional layers, taking an absolute value of a result and then passing through the global mean pooling layer, then passing through the fully connected layer, multiplying a fully connected output by an input to obtain a soft threshold, and then adding into the global residual for output.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of ZHANG of having wherein in the second step, the third image is generated based on a second residual image generated from the first residual image using soft thresholding.
Wherein SU’s image processing method wherein in the second step, the third image is generated based on a second residual image generated from the first residual image using soft thresholding.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and restoration details, since both SU and ZHANG relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and ZHANG relates to a single image dehazing method based on detail recovery; while effectively removing fog, it can retain more image detail information, reduce the blurring at the edges of the defogged image, and generate higher-quality defogged images. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and ZHANG (US 20240289928 A1), Paragraph [0005].
Regarding claim 6, SU and SHROERS in view of ZHANG teach the image processing method according to claim 5,
SU and SHROERS fail to explicitly teach wherein in the second step, the second residual image is generated by applying a coefficient larger than 1 based on a threshold value of the soft thresholding to a residual image obtained by performing the soft thresholding for the first residual image.
However, ZHANG explicitly teaches wherein in the second step, the second residual image is generated (Fig. 3, Paragraph [0041] – ZHANG discloses the introduced feature enhancement module includes three residual blocks and one global residual, by adding a previous dehazing image to an input image with fog to enhance signals, dehazing the enhanced image, and subtracting the previous dehazing image from a restored signal enhancement result to obtain an image with better signal-to-noise ratio and better recovery features, and there is also a skip connection between the encoding region and the decoding region.)
by applying a coefficient larger than 1 based on a threshold value of the soft thresholding to a residual image obtained by performing the soft thresholding for the first residual image (Fig. 5, Paragraph [0044] – ZHANG discloses after two convolutions of the input features, the residual shrinkage module takes the absolute values of all the features of the feature image, which are then pooled and averaged with the global mean to obtain a feature A. In another path, the feature image after global mean pooling is input to a small fully connected network that takes the Sigmoid function as the last layer, then the output is normalized to between 0 and 1 to obtain a coefficient α, such that the final threshold can be expressed as A×α [wherein coefficient α can be a coefficient larger than 1]. Different thresholds for different samples allow features that are not relevant to the current task to be set to zero by soft thresholding, and features that are relevant to the current task to be kept.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of ZHANG of having wherein in the second step, the second residual image is generated by applying a coefficient larger than 1 based on a threshold value of the soft thresholding to a residual image obtained by performing the soft thresholding for the first residual image.
Wherein SU’s image processing method wherein in the second step, the second residual image is generated by applying a coefficient larger than 1 based on a threshold value of the soft thresholding to a residual image obtained by performing the soft thresholding for the first residual image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and restoration details, since both SU and ZHANG relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and ZHANG relates to a single image dehazing method based on detail recovery; while effectively removing fog, it can retain more image detail information, reduce the blurring at the edges of the defogged image, and generate higher-quality defogged images. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and ZHANG (US 20240289928 A1), Paragraph [0005].
Regarding claim 7, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein in the second step, the third image is generated based on a weight map based on the absolute value of the first residual image, the fourth image, and one of the first residual image and the second image.
However, ZHANG explicitly teaches wherein in the second step, the third image is generated (Fig. 1, Paragraph [0048] – ZHANG discloses a final output image is the defogged image.)
based on a weight map based on the absolute value of the first residual image (Fig. 5, Paragraph [0043] – ZHANG discloses the residual shrinkage module includes two 3×3 convolutional layers, a global mean pooling layer, and a fully connected layer, inputting feature images into the two 3×3 convolutional layers, taking an absolute value of a result and then passing through the global mean pooling layer, then passing through the fully connected layer, multiplying a fully connected output by an input to obtain a soft threshold, and then adding into the global residual for output. Paragraph [0054] – ZHANG discloses an output feature of the residual block is put through one ReLU activation function first, then through two convolution layers and a sigmoid function to change into a 1×H×W output feature, and the 1×H×W output feature represents the weight information of each pixel point. Paragraph [0056] – ZHANG discloses the input F* is multiplied with its corresponding pixel weights PA to achieve the assignment of different weights to different pixels.),
the fourth image (Fig. 3, Paragraph [0041] – ZHANG discloses the introduced feature enhancement module includes three residual blocks and one global residual, by adding a previous dehazing image to an input image [wherein an input image is a fourth image] with fog to enhance signals, dehazing the enhanced image, and subtracting the previous dehazing image from a restored signal enhancement result to obtain an image with better signal-to-noise ratio and better recovery features.),
and one of the first residual image and the second image (Fig. 5, Paragraph [0043] – ZHANG discloses the residual shrinkage module includes two 3×3 convolutional layers, a global mean pooling layer, and a fully connected layer, inputting feature images into the two 3×3 convolutional layers, taking an absolute value of a result and then passing through the global mean pooling layer, then passing through the fully connected layer, multiplying a fully connected output by an input to obtain a soft threshold, and then adding into the global residual for output.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of ZHANG of having wherein in the second step, the third image is generated based on a weight map based on the absolute value of the first residual image, the fourth image, and one of the first residual image and the second image.
Wherein SU’s image processing method wherein in the second step, the third image is generated based on a weight map based on the absolute value of the first residual image, the fourth image, and one of the first residual image and the second image.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and restoration details, since both SU and ZHANG relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and ZHANG relates to a single image dehazing method based on detail recovery; while effectively removing fog, it can retain more image detail information, reduce the blurring at the edges of the defogged image, and generate higher-quality defogged images. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and ZHANG (US 20240289928 A1), Paragraph [0005].
Regarding claim 8, SU and SHROERS in view of ZHANG teach the image processing method according to claim 7,
SU in view of SHROERS fail to explicitly teach wherein the weight map represents a weight of the first residual image or the second image, and wherein in the weight map, the weight reduces for a pixel in the first residual image having a smaller non-zero absolute value.
However, ZHANG explicitly teaches wherein the weight map represents a weight of the first residual image or the second image (Fig. 1, Paragraph [0054] – ZHANG discloses an output feature of the residual block is put through one ReLU activation function first, then through two convolution layers and a sigmoid function to change into a 1×H×W output feature, and the 1×H×W output feature represents the weight information of each pixel point. Paragraph [0056] – ZHANG discloses the input F* is multiplied with its corresponding pixel weights PA to achieve the assignment of different weights to different pixels.),
and wherein in the weight map, the weight reduces for a pixel in the first residual image having a smaller non-zero absolute value (Fig. 2, Paragraph [0041] – ZHANG discloses the improved residual module includes: passing input features through a basic residual block, which are then adding global residual as a whole through a pixel attention mechanism. By combining the pixel attention mechanism to assign different weights to different pixels, the more interested pixels get more attention for better feature extraction. Paragraph [0044] – ZHANG further discloses after two convolutions of the input features, the residual shrinkage module takes the absolute values of all the features of the feature image, which are then pooled and averaged with the global mean to obtain a feature A.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of ZHANG of having wherein the weight map represents a weight of the first residual image or the second image, and wherein in the weight map, the weight reduces for a pixel in the first residual image having a smaller non-zero absolute value.
Wherein SU’s image processing method wherein the weight map represents a weight of the first residual image or the second image, and wherein in the weight map, the weight reduces for a pixel in the first residual image having a smaller non-zero absolute value.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and restoration details, since both SU and ZHANG relate to image processing and enhancement techniques, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and ZHANG relates to a single image dehazing method based on detail recovery; while effectively removing fog, it can retain more image detail information, reduce the blurring at the edges of the defogged image, and generate higher-quality defogged images. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and ZHANG (US 20240289928 A1), Paragraph [0005].
Claims 10 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS in further view of LITWILLER (US 20210272240 A1), hereinafter referenced as LITWILLER.
Regarding claim 10, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein in the first step, the input data input to the machine learning model includes the first image and a first map, and wherein the first map has a plurality of channels.
However, LITWILLER explicitly teaches wherein in the first step, the input data input to the machine learning model includes the first image and a first map (Fig. 2, Paragraph [0040] – LITWILLER discloses inputs, which may include noise parameters 204 [wherein noise parameters 204 are a first map], may be propagated through the plurality of layers within deep neural network 224, to map intensity values of noisy medical image 206 [wherein noisy medical image 206 is the first image] to intensity values of de-noised image 220.),
and wherein the first map has a plurality of channels (Fig. 2, Paragraph [0040] – LITWILLER discloses noise parameters 204 may each be incorporated into the deep neural network 324 via a plurality of distinct mechanisms/channels.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of LITWILLER of having wherein in the first step, the input data input to the machine learning model includes the first image and a first map, and wherein the first map has a plurality of channels.
Wherein SU’s image processing method wherein in the first step, the input data input to the machine learning model includes the first image and a first map, and wherein the first map has a plurality of channels.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and quality, since both SU and LITWILLER relate to image processing techniques using neural networks, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and LITWILLER enables a deep neural network to be trained to identify colored noise in medical images, and to produce de-noised images with substantially less colored noise, thus increasing image resolution and diagnostic quality. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and LITWILLER (US 20210272240 A1), Paragraph [0086].
Regarding claim 11, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein in the first step, input data input to the machine learning model includes the first image and a first map, and wherein the first map has nonuniform values in a vertical or horizontal direction.
However, LITWILLER explicitly teaches wherein in the first step, input data input to the machine learning model includes the first image and a first map (Fig. 2, Paragraph [0040] – LITWILLER discloses inputs, which may include noise parameters 204 [wherein noise parameters 204 are a first map], may be propagated through the plurality of layers within deep neural network 224, to map intensity values of noisy medical image 206 [wherein noisy medical image 206 is the first image] to intensity values of de-noised image 220.),
and wherein the first map has nonuniform values in a vertical or horizontal direction (Fig. 2, Paragraph [0038] – LITWILLER discloses noise parameters 204 includes acquisition data for the noisy medical image 206. LITWILLER further discloses acquisition data may further include whether post-processing (e.g., such as image sharpening) has been applied to noisy medical image 206. In particular, the k-space sampling pattern and the k-space sampling density may impact the type(s) and distribution(s) of colored noise in a medical image, especially when the k-space sampling pattern is characterized by nonuniform oversampling or nonuniform undersampling.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of LITWILLER of having wherein in the first step, input data input to the machine learning model includes the first image and a first map, and wherein the first map has nonuniform values in a vertical or horizontal direction.
Wherein SU’s image processing method wherein in the first step, input data input to the machine learning model includes the first image and a first map, and wherein the first map has nonuniform values in a vertical or horizontal direction.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and quality, since both SU and LITWILLER relate to image processing techniques using neural networks, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and LITWILLER enables a deep neural network to be trained to identify colored noise in medical images, and to produce de-noised images with substantially less colored noise, thus increasing image resolution and diagnostic quality. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and LITWILLER (US 20210272240 A1), Paragraph [0086].
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in further view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS in further view of QIAN (US 20210370993 A1), hereinafter referenced as QIAN.
Regarding claim 12, SU in view of SHROERS teach the image processing method according to claim 1,
Although SU further teaches wherein the machine learning model has a plurality of model parameters previously determined by training (Fig. 3, Paragraph [0091] – SU discloses during neural network model training (e.g., 108), the training code can extract the training data and use such data to adjust the model parameters (e.g., weights used by convolutional layers of the NN model 306) for performing UR processing. This speeds up the model training by avoiding complicated image processing on the fly and allows the model to be trained repeatedly with different parameters associated with the training data.),
SU in view of SHROERS fail to explicitly teach and wherein the number of bits of at least one of the plurality of model parameters is smaller during generating the first residual image than during the training.
However, QIAN explicitly teaches and wherein the number of bits of at least one of the plurality of model parameters is smaller during generating the first residual image than during the training (Fig. 2, Paragraph [0143] – QIAN discloses track inspection networks are typically trained using 32-bit floating point arithmetic, but weight quantization allows us to reduce our model parameters to 8-bit representations without sacrificing accuracy. See also Fig. 24, Paragraph [0175].).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of QIAN of having and wherein the number of bits of at least one of the plurality of model parameters is smaller during generating the first residual image than during the training.
Wherein SU’s image processing method wherein having and wherein the number of bits of at least one of the plurality of model parameters is smaller during generating the first residual image than during the training.
The motivation behind this modification would have been to provide an enhanced image processing method that can improve image resolution and processing speed, since both SU and QIAN relate to image processing techniques using neural networks, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and QIAN discloses a real-time pixel level detection system with an improved real-time instance segmentation model that leverages the speed of real-time object detection and high accuracy of two-stage instance segmentation; fast processing speeds up inspection videos into real-time useful information as the videos are being recorded to assist track maintenance decisions. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and QIAN (US 20210370993 A1), Paragraph [0088].
Claims 15 is rejected under 35 U.S.C. 103 as being unpatentable over SU (US 20200402205 A1), hereinafter referenced as SU in further view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS in further view of BIJALWAN (US 20230281423 A1), hereinafter referenced as BIJALWAN.
Regarding claim 15, SU in view of SHROERS teach the image processing method according to claim 1,
SU in view of SHROERS fail to explicitly teach wherein the machine learning model has a plurality of first layers having a first bit number and a plurality of second layers having a second bit number smaller than the first bit number.
However, BIJALWAN explicitly teaches wherein the machine learning model has a plurality of first layers having a first bit number (Fig. 1, Paragraph [0038] – BIJALWAN discloses quantization module 118 quantizes each group of layers selected by the grouping module 116 into high precision and/or lower precision format to generate a final mixed precision model. Paragraph [0078] – BIJALWAN further discloses the quantization module 118 clusters the layers corresponding to a range difference values between 5 bits-7 bits to a first cluster, the layers corresponding to a range of difference values between 4 bits-5 bits to a second cluster and so on.)
and a plurality of second layers having a second bit number smaller than the first bit number (Fig. 1, Paragraph [0038] – BIJALWAN discloses quantization module 118 quantizes each group of layers selected by the grouping module 116 into high precision and/or lower precision format to generate a final mixed precision model. Paragraph [0078] – BIJALWAN further discloses the quantization module 118 clusters the layers corresponding to a range difference values between 5 bits-7 bits to a first cluster, the layers corresponding to a range of difference values between 4 bits-5 bits to a second cluster and so on.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of SU in view of SHROERS of having image processing method comprising: a first step of generating a first residual image based on a first image using a machine learning model; and a second step of generating a third image based on a fourth image and an absolute value of the first residual image, with the teachings of BIJALWAN of having wherein the machine learning model has a plurality of first layers having a first bit number and a plurality of second layers having a second bit number smaller than the first bit number.
Wherein SU’s image processing method wherein the machine learning model has a plurality of first layers having a first bit number and a plurality of second layers having a second bit number smaller than the first bit number.
The motivation behind this modification would have been to provide an enhanced image processing method that can provide improved image resolution and image compression rate, since both SU and BIJALWAN relate to image processing techniques using neural networks, wherein SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing, and BIJALWAN relates to a method and system for quantization aware training of a neural network for image compression that improves accuracy of a trained neural network model and improves image compression rate compared to a high precision model. Please see SU (US 20200402205 A1), Paragraph [0041, 0092], and BIJALWAN (US 20230281423 A1), Paragraph [0002, 0032].
Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over JANGID (US 20240062342 A1), hereinafter referenced as JANGID in view of SU (US 20200402205 A1), hereinafter referenced as SU in further view of SHROERS (US 20210304387 A1), hereinafter referenced as SHROERS.
Regarding claim 18, JANGID teaches an image pickup apparatus (Fig. 1, #100 called network configuration, Paragraph [0033]) comprising:
an image processing apparatus (Fig. 1, #101 called electronic device, Paragraph [0047] – JANGID discloses FIG. 2 illustrates an example image processing pipeline 200 in an electronic device in accordance with this disclosure. For ease of explanation, the image processing pipeline 200 shown in FIG. 2 is described as being implemented on or supported by the electronic device 101 in the network configuration 100 of FIG. 1.);
Although JANGID further teaches and an image sensor (Fig. 1, #180 called sensor(s), Paragraph [0041]) configured to acquire a captured image as the first image (Fig. 1, Paragraph [0041] – JANGID discloses electronic device 101 further includes one or more sensors 180 that can meter a physical quantity or detect an activation state of the electronic device 101 and convert metered or detected information into an electrical signal. For example, one or more sensors 180 include one or more cameras or other imaging sensors, which may be used to capture images of scenes.);
JANGID fails to explicitly teach wherein the image processing apparatus includes: a processor configured to generate a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image,
However, SU explicitly teaches wherein the image processing apparatus (Fig. 9, #900 called computer, Paragraph [0109-0110] – SU discloses FIG. 9 is a block diagram illustrating circuitry for a device that implements algorithms and performs methods, according to example embodiments. One example computing device in the form of a computer 900.) includes:
a processor (Fig. 9, #905 called processor, Paragraph [0110] – SU discloses computer 900 (also referred to as computing device 900, computer system 900, or computer 900) may include a processor 905.)
configured to generate a first residual image based on a first image (Fig. 3, Paragraph [0076] – SU discloses the NN model 306 uses multiple convolutional layers to involve the LR residual image 304 and generate a plurality of HR residual sub-images 308 corresponding to the input LR image 302 [wherein input LR image 302 is a first image]. In some aspects, the plurality of HR residual sub-images 308 includes four sub-images. The pixel shifting module 318 performs pixel shifting on the plurality of HR residual sub-images 308 to generate an HR residual image 310 [wherein HR residual image 310 is a first residual image].)
using a machine learning model (Fig. 3, #306 called NN model, Paragraph [0017] – SU discloses the neural network model includes an input layer and the plurality of convolutional layers comprises four convolutional layers. The input layer is configured to receive the digital image. An output layer of the four convolutional layers is configured to output the plurality of HR residual sub-images. Paragraph [0078] – SU further discloses one or more functionalities performed by the bicubic upsampling module 316, the residue generation module 322, the pixel shifting module 318, and the adder 320 can be performed by the NN model 306.);
wherein the fourth image is the first image or an image based on the first image (Fig. 3, Paragraph [0076] – SU discloses a low-resolution (LR) input image 302 [wherein input LR image 302 is the first image] is processed (e.g., by the bicubic upsampling module 316 and the residue generation module 322) to generate a base HR image 312 [wherein base HR image 312 is the fourth image].),
and wherein in a case where a second image is a sum of the fourth image and the first residual image (Fig. 3, Paragraph [0076] – SU discloses adder 320 may comprise suitable circuitry, logic, interfaces, or code and is configured to add the base HR image 312 [wherein base HR image 312 is the fourth image] with the HR residual image 310 [wherein HR residual image 310 is the first residual image] to generate an HR image 314 [wherein HR image 314 is a second image] as an output image corresponding to the input LR image 302.),
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of JANGID of having an image pickup apparatus comprising: an image processing apparatus; and an image sensor configured to acquire a captured image as the first image; with the teachings of SU of having wherein the image processing apparatus includes: a processor configured to generate a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image.
Wherein JANGID’s image pickup apparatus wherein the image processing apparatus includes: a processor configured to generate a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image.
The motivation behind this modification would have been to provide an enhanced image processing technique that reduces complexity and improves image resolution, since both JANGID and SU relate to image processing and enhancement techniques, wherein JANGID relates to machine learning-based approaches for synthetic training data generation and image sharpening, resulting in a low-complexity AI sharpening network that is highly effective in removing blur from images, and SU relates to changing the image resolution of images using a neural network wherein the use of a concise neural network model that combines the functions of super-resolution, noise reduction, removal of blocking artifacts, and local contrast enhancement, the concise neural network model can be deployed on mobile devices for performing real-time UR processing. Please see JANGID (US 20240062342 A1), Paragraph [0001, 0078], and SU (US 20200402205 A1), Paragraph [0041, 0092].
JANGID in view of SU fail to explicitly teach and generate a third image based on a fourth image and an absolute value of the first residual image, the processor makes sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
However, SHROERS explicitly teaches and generate a third image (Fig. 4, Paragraph [0052] – SHROERS discloses the method outlined by flowchart 460 may continue and conclude with optionally correcting input image 230/330/530 [wherein the corrected input image is a third image] using a respective one of inpainted masked patches 354a-354l corresponding to the location of the one or more anomalous pixels in input image 230/330/530 that were identified in action 465 (action 468).)
based on a fourth image image (Fig. 4, Paragraph [0049] - SHROERS discloses flowchart 460 continues with identifying one or more anomalous pixels in input image 230/330/530 [wherein input image is a fourth image] using residual image 236/536 (action 465) [wherein residual image is the first residual image].)
and an absolute value of the first residual image (Fig. 4, Paragraph [0039] – SHROERS further discloses pixels in residual image 236 may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330.),
the processor (Fig. 1, #104 called hardware processor, Paragraph [0014]) makes sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image (Fig. 4, Paragraph [0042] – SHROERS discloses identification of those bright pixels as one or more anomalous pixels may be performed by software code 110, executed by hardware processor 104. Action 465 may include detecting one or more anomaly candidates in input image 230/330/530 using residual image 236/536, based, for example, on the brightness of the one or more anomaly candidates, and determining a residual value associated with each of the one or more anomaly candidates, such as a residual value corresponding to its relative brightness. Paragraph [0039] – SHROERS further discloses where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch.)
as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases (Fig. 2, Paragraph [0039] – SHROERS discloses pixels in residual image 236 [wherein residual image 236 is a first residual image] may each have an intensity corresponding to the absolute value resulting from subtraction of the inpainted one or more masked patches predicted in action 463 from the one or more patches 354a-354l in original input image 230/330. Thus, where pixels in an inpainted masked patch match corresponding pixels in original input image 230/330, corresponding pixels in residual image 236 can be expected to appear dark, while pixels in residual image 236 where pixels in inpainted masked patches fail to match pixels in original input image 230/330 may have a brightness proportional to the mismatch. [See also Fig. 4.]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date the claimed invention was made to combine the teachings of JANGID in view of SU of having an image pickup apparatus comprising: an image processing apparatus; and an image sensor configured to acquire a captured image as the first image; wherein the image processing apparatus includes: a processor configured to generate a first residual image based on a first image using a machine learning model; wherein the fourth image is the first image or an image based on the first image, and wherein in a case where a second image is a sum of the fourth image and the first residual image, with the teachings of SHROERS of having and generate a third image based on a fourth image and an absolute value of the first residual image, the processor makes sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
Wherein JANGID’s image pickup apparatus wherein having and generate a third image based on a fourth image and an absolute value of the first residual image, the processor makes sets a signal value of each pixel in the third image to a value closer to a signal value of a pixel in the fourth image corresponding to each pixel in the third image than a signal value of a pixel in the second image corresponding to each pixel in the third image as a non-zero absolute value of a signal value of a pixel in the first residual image corresponding to each pixel in the third image decreases.
The motivation behind this modification would have been to provide an enhanced image processing technique that reduces complexity and automates pixel error detection, since both JANGID and SHROERS relate to image processing and enhancement techniques, wherein JANGID relates to machine learning-based approaches for synthetic training data generation and image sharpening, resulting in a low-complexity AI sharpening network that is highly effective in removing blur from images, and SHROERS relates to systems and methods for performing automated pixel error detection using an inpainting neural network, the present solution advantageously yields better results than other attempts to automate pixel error detection, with lower false positive rates and very high recall, and in fact enables the detection of pixel errors that human inspectors and previous automated pixel error detection solutions had failed to identify. Please see JANGID (US 20240062342 A1), Paragraph [0001, 0078], and SHROERS (US 20210304387 A1), Paragraph [0003, 0056].
Conclusion
Listed below are the prior arts made of record and not relied upon but are considered pertinent to applicant’s disclosure.
NOVIKOV et al. (US 20240233093 A1) - Exemplary systems, methods and computer arrangement for decorrelating and removing noise, and/or for removing Gibbs ringing data from at least one image can be provided. In one example, there can be a procedure for receiving information related to the at least one image; a procedure for producing the at least one image based on the determination if the noise is decorrelated; a procedure to remove the noise from the at least one image; a procedure for detecting an oscillation pattern from at least one edge of a measured portion of k-space associated with the information; and a procedure for removing the Gibbs ringing artifact from the information based on the detected oscillation pattern......… Fig. 1, Abstract.
SCHNELLBACHER et al. (US 20240095885 A1) - A mechanism for generating a partially denoised image. A residual noise image, obtained by processing an image using a convolutional neural network, is weighted. The blending or combination of the weighted residual noise image and the (original) image generates the partially denoised image.........… Fig. 1, Abstract.
LI et al. (US 20220398698 A1) - An image processing method includes obtaining a to-be-processed image, performing an image processing operation on the to-be-processed image by inputting the to-be-processed image into a corresponding image processing model to obtain a processed image, and obtaining an output image according to the processed image of the corresponding image processing model. The image processing operation includes at least one of color deviation removal processing or ghost effect removal processing. The image processing model corresponding to the color deviation removal processing is a first image processing model, and the image processing model corresponding to the ghost effect removal processing is a second image processing model.........… Fig. 1, Abstract.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BEZAWIT N SHIMELES whose telephone number is (571)272-7663. The examiner can normally be reached M-F 7:30am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Chineyere Wills-Burns can be reached at (571) 272-9752. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BEZAWIT NOLAWI SHIMELES/Examiner, Art Unit 2673
/CHINEYERE WILLS-BURNS/Supervisory Patent Examiner, Art Unit 2673
Prosecution Insights