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 .
Information Disclosure Statement
The information disclosure statement (IDS) were filed on 11/21/2024 and 10/31/2025. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 7 and 14 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “long-exposure” in claims 7 and 14 is a relative term which renders the claim indefinite. The term “long-exposure” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The term “long-exposure” renders the following limitations in claims 7 and 14 as indefinite; “the training image frames capture different static scenes; the ground truth images comprise long-exposure images of the static scenes; and the additional sets of training image frames simulate inter-frame motion and inter-frame misalignment.” One of ordinary skill in the art would not be able to determine what are the metes and bounds of “long-exposure” since it is a subjective term. One of ordinary skill in the art would ask: “What can exactly be considered long? 1 second, 1 minute, 1 day?”. Therefore one of ordinary skill in the art would not be able to apprise the scope of the claim for reasons regarding clarity.
Claim Rejections - 35 USC § 103
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, 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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 7, 8, 14 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Hu et. al. (US Pub. No. 20200265567 A1) in view of Li et. al. (US Pub. No. 20230252608 A1).
As per claim 1, Hu teaches “A method comprising:
obtaining, using at least one processing device of an electronic device, multiple sets of training image frames, each set of training image frames having an associated ground truth image;” (See paragraph 76 “[0076] The training of convolutional neural networks usually requires the use of a large number of training examples. To support convolutional neural network-based multi-exposure fusion of multiple image frames, each training example could include a set of LDR image frames of a dynamic scene (or other scene) and “ground truth” blending maps associated with the image frames. ” See also paragraphs 80-95. Hu)
“applying, using the at least one processing device… and warping to the multiple sets of training image frames in order to generate additional sets of training image frames; and” (See paragraph 88 “[0088] One or both of these approaches can also be performed for the images and ground truth blending maps of the validation set. For example, assuming the validation set includes 40 sets of images, the techniques shown in FIGS. 4 and 5 can be used to produce 400 additional sets of image training patches and associated ground truth training patches. These can be supplemented with synthetic image training patches and associated ground truth training patches that are generated using the techniques shown in FIGS. 6 and 7 if needed or desired.”, the techniques presented include those of paragraphs 77-87, such as warping to create a sense of motion (which is interpreted as motion blur) “[0085]… These images 602, 604, and 606 are processed to generate a set 608 of synthesized images 610, 612, and 614. In order to generate the synthesized image set 608, a warping operator 616 is applied to the image 602, and a warping operator 618 is applied to the image 606. Each warping operator 616 and 618 generally represents motion objects with random shapes and motion vectors. Applying the warping operator 616 to the image 602 creates the appearance of motion when comparing the images 602 and 610, and applying the warping operator 618 to the image 606 creates the appearance of motion when comparing the images 606 and 614. Thus, it is possible to artificially create the appearance of motion using images of a static scene, effectively converting images of the static scene into images of a dynamic scene with known motion...” Hu)
“training, using the at least one processing device, a machine learning model to align image frames and remove motion blur from the image frames based on at least the additional sets of training image frames and the ground truth images.” (They are aligned as seen on paragraphs 57-61 and 112-114 “[0112]…The images are aligned and otherwise pre-processed at step 1204. This could include, for example, the processor 120 of the electronic device 101 aligning the image frames 204, 206, and 208 by performing the image registration operation 210… [0114]… [0114] Image frames are synthesized based on the aligned image frames to remove motion from the aligned image frames at step 1208.” The motion blur is removed as seen on paragraphs 139, 149-150 and 156-159 “[0156] The additional images are aligned and pre-processed at step 2408. This could include, for example, the processor 120 of the electronic device 101 selecting one of the additional image frames as a reference frame and aligning the non-reference image frames to the reference frame both geometrically and photometrically. Motion maps identifying portions of the aligned image frames that are prone to blur are generated at step 2410… This could include, for example, the processor 120 of the electronic device 101 performing the blending operation 1710 to combine the aligned image frames based on the motion maps.” “[0157]… [0157] The blended image frame is deblurred using the motion maps at step 2414…”. See also Paragraphs 38-39 57-58, 61 and 112-120. Hu), however Hu does not completely teach “applying… motion blur… to generate additional sets of training image frames…”
Li teaches “applying… motion blur… to generate additional sets of training image frames…” (See paragraph 43 “[0043] Embodiments of the present disclosure describe various techniques to create training data to train an AI-based image processing operation… For each image frame captured during the multi-frame capture operation, a motion-distorted image frame is generated. The generated motion-distorted image frames can be combined using an MFP operation to generate a ground truth image, which allows the ground truth image to contain known motion blur…” See also paragraphs 4-5 and 40-43. Li)
It would have been obvious to one of ordinary skill in the art before the effective filing
date of the claimed invention to combine the teachings of Hu with the teachings of Li to also apply motion blur to generate additional training images. The modification would have been motivated by the desire to produce higher quality images, therefore it is an improvement, as suggested by Li (See paragraphs 40-43 “[0040] As another particular example, an AI-based multi-frame processing (MFP) operation enables an electronic device to capture and combine multiple image frames in order to produce higher-quality images. Image quality improvements, such as HDR, low-light photography, and motion blur reduction, are enabled based on MFP operations. In order to train an AI-based MFP operation, multiple image pairs are often required, where each pair includes an input image and a ground truth image. If the input image is a handheld image and the ground truth image is a stationary image, an AI network can be trained to remove blur due to handshake motion when combining the images together… [0043]… Once trained, the AI-based image processing operation can be used to remove small camera motions due to handshake in an MFP operation.” Li)
Claim 8 is rejected under the same analysis as claim 1.
Claim 15 is rejected under the same analysis as claim 1. (Paragraphs 38-39 57-58, 61 and 112-120 shows alignment of inputted frames of a scene to reduce motion blur in images to create a final image. )
As per claim 7, Hu in view of Li teaches “the method of Claim 1, wherein: the training image frames capture different static scenes; the ground truth images comprise long-exposure images of the static scenes; and the additional sets of training image frames simulate inter-frame motion and inter-frame misalignment.” (See paragraphs 83-87 “[0085]… [0085]… The images 602, 604, and 606 are of the same scene and are captured using different camera exposures. For example, the images 602, 604, and 606 could be captured with exposure biases of {−2.0, +0.0, +1.0}, respectively, although other camera exposures could be used. In this example, the scene is substantially or completely static, meaning there is very little or no movement in the scene. These images 602, 604, and 606 are processed to generate a set 608 of synthesized images 610, 612, and 614. In order to generate the synthesized image set 608, a warping operator 616 is applied to the image 602, and a warping operator 618 is applied to the image 606. Each warping operator 616 and 618 generally represents motion objects with random shapes and motion vectors. Applying the warping operator 616 to the image 602 creates the appearance of motion when comparing the images 602 and 610, and applying the warping operator 618 to the image 606 creates the appearance of motion when comparing the images 606 and 614. Thus, it is possible to artificially create the appearance of motion using images of a static scene, effectively converting images of the static scene into images of a dynamic scene with known motion... [0087] Because the images 602, 604, and 606 are taken of a static scene, the ground truth blending map 710 generated between the images 602 and 610 also (ideally) represents the same ground truth blending map 710 between the image 604 (i.e. image 612) and the image 610 since the image 612 is not warped.” )
(Li shows the use of longer exposure for images with a clear obvious rationale for improvement, see paragraphs 60 and 68 “0068] In FIG. 2B, an image frame 270 may represent one of the image frames 202 captured using a longer exposure and a tripod or other mechanism to safeguard against motion… The image frame 274 is blurry due to the electronic device being handheld and is noisy due to a high ISO value (compared to the image frame 272)… The similarity between the image frame 274 and the image frame 276 provides an indication that the synthetic multi-frame capture operation 210 can be used to successfully simulate image capture using a handheld electronic device, and this can be achieved without the need for manually capturing multiple image frames in order to train an AI-based image processing operation.” It also simulates inter frame misalignment as seen on paragraph 81 and figs. 3A-3E. Li)
Claim 14 is rejected under the same analysis as claim 7.
Claims 2-4, 9-11 and 16-18 are rejected under 35 U.S.C. 103 as being unpatentable over Hu in view of Li and further in view of Watson et. al. (WO Pub. No. 2024163522 A1) .
As per claim 2, Hu in view of Li already teaches “The method of Claim 1, wherein applying the motion blur comprises, for each set of training image frames… associated ground truth image” however Hu in view of Li does not completely teach “identifying noise in the associated ground truth image; removing the identified noise from the training image frames in the set of training image frames in order to generate denoised image frames; applying one or more random blur kernels to each of the denoised image frames in order to generate blurred image frames; and adding the identified noise to the blurred image frames.”
Watson teaches “identifying noise in the associated ground truth image; removing the identified noise from the training image frames in the set of training image frames in order to generate denoised image frames;” (See paragraphs 65-66 “[0065 ] Image noise 320 may represent noise identified in noise image 304 or intermediate noise image 324 (depending on the iteration) by noise model 318… [0066] Additionally, at each iteration performed by diffusion model 312, denoised image calculator 322 may be configured to determine intermediate noise image 324. Specifically, denoised image calculator 322 may subtract image noise 320 from intermediate noise image 324 based on which noise model 318 determined image noise 320 (i.e., the instance of intermediate noise image 324 from the preceding iteration)…” “[0077] Training noise conditioning value 410 may indicate and/or control an amount of noise to be added by degradation calculator 412 to target image 402 when generating training conditioning image 414.” See also fig. 3. Watson ) “applying one or more random blur kernels to each of the denoised image frames in order to generate blurred image frames; and adding the identified noise to the blurred image frames.” (See fig. 3 and fig. 5, paragraphs 75-77, 79-84 and 89-99. “[0075]… To determine training conditioning image 414, degradation calculator 412 may be configured to blur, resize, compress, remove parts of, and/or add noise to target image 402… [0079] Forward diffusion calculator 404 may be configured to determine noisy target image 406 based on target image 402 and forward process noise 416. Specifically, forward diffusion calculator 404 may be configured to determine forward process noise 416 (e.g., Gaussian noise) and add it to target image 402 using a forward diffusion process that diffusion model 312 is being trained to reverse…” “[0090] Each of blur calculators 504-506 may be configured to blur an image provided thereto as input. Each of blur calculators 504-506 may include a plurality of blur filters… [0091] For example, each of blur calculators 504-506 may include a Gaussian filter, a generalized Gaussian filter, a plateau-based kernel filter, and a sine filter, each of which maybe selected with a corresponding probability (e.g., 0.63, 0.135, 0.135 and 0.1, respectively)… Anisotropic filters may be rotated by a random angle (e.g., randomly selected from (— n, ??]). A radius of each of the filters may be varied (e.g., randomly) between, for example, 3 and 11 pixels. The filters may have a standard deviation that is selected (e.g., randomly) from a predetermined range (e.g., from [0.2,3.0], or [0.2, 1.5])…” Watson)
It would have been obvious to one of ordinary skill in the art before the effective filing
date of the claimed invention to combine the teachings of Hu with the teachings of Li and Watson to denoise images and apply random blur kernels to generate blur images and adding the noise onto the image when applying motion blur. The modification would have been motivated by the desire to have higher quality images through enhancement and better generalization in order to produce more accurate results , therefore it is an improvement, as suggested by Watson (See paragraphs 89, 92 and 96 “[0096] Applying degradation process 500-502 to target image 402 may allow' diffusion model 312 to learn to enhance both in-distribution images and out-of-distribution images. For example, the variations in how degradation processes 500-502 modify target image 402 may expose diffusion model 312 to a wide range of potential image deformations that may be representative of deformations that diffusion model 312 is likely to encounter at inference time, including in images that differ from the target images used at training time. That is, training diffusion model 312 on the basis of degradation processes 500-502 may allow diffusion model 312 to better generalize to, and thus produce more accurate results with respect to, images that differ from the target images used at training time.” Watson)
Claim 9 is rejected under the same analysis as claim 2.
Claim 16 is rejected under the same analysis as claim 2.
As per claim 3, Hu in view of Li and Watson already teaches “The method of Claim 2, wherein: each of the training image frames comprises image data in multiple color channels; and the one or more random blur kernels are applied to each color channel of each training image frame.” (Hu already teaches image data comprising multiple color channels, see paragraphs 71, 79, 81 and 113 “[0079]… Each of the images 402, 404, and 406 could be associated with multiple color channels, such as three color channels” along with applying convolutions/filters/kernels on each color channel, “[0071] In some embodiments, each input image patch 302 includes multiple color “channels,” each of which typically represents one color contained in the associated image patch 302. For example, digital cameras often support red, green, and blue color channels. By concatenating the input image patches 302 along the color channels, the number of inputs to the convolutional layer 310a can be increased. For example, if there are M input image patches 302 each with N color channels, concatenating the inputs along the color channels can produce M×N inputs to the convolutional layer 310a. Similarly, the convolutional layer 316d can generate feature maps having M×N weight channels, and the convolutional layer 318 can process the feature maps to generate M blending map patches 304.” And Watson already teaches random blur kernels applied on paragraph 91. )
Claim 10 is rejected under the same analysis as claim 3.
Claim 17 is rejected under the same analysis as claim 3.
As per claim 4, Hu in view of Li and Watson already teaches “The method of Claim 2, wherein applying the one or more random blur kernels to each of the denoised image frames comprises: selecting an orientation and strength of motion to be created in each of the denoised image frames; and defining the one or more random blur kernels for each of the denoised image frames based on the corresponding orientation and strength of motion.” (See paragraphs 89-91 in reference Hu. “[0091] For example, each of blur calculators 504-506 may include a Gaussian filter, a generalized Gaussian filter, a plateau-based kernel filter, and a sine filter, each of which maybe selected with a corresponding probability (e.g., 0.63, 0.135, 0.135 and 0.1, respectively). Various properties of each filter may be selected according to corresponding probabilities. For example, the Gaussians may be isotropic with probability 9/14 and anisotropic otherwise, and/or the plateau kernel may be isotropic with probability 0.8 and anisotropic otherwise. Anisotropic filters may be rotated by a random angle (e.g., randomly selected from (— n, ??]). A radius of each of the filters may be varied (e.g., randomly) between, for example, 3 and 11 pixels. The filters may have a standard deviation that is selected (e.g., randomly) from a predetermined range (e.g., from [0.2,3.0], or [0.2, 1.5]). For the sine filter, the cutoff frequency may be selected (e.g., randomly) from, for example, [rr/3, TT] when the radius is less than 6 pixels and from [TF/5, TT] otherwise. The shape parameter /? may be sampled from, for example, [0.5, 4.0] for the generalized Gaussian filter and from [1.0, 2.0] for the plateau filter. Application of blur calculator 506 may be omitted with probability 0.2.” I is well known in the art that the radius/standard deviation represent a strength of blurring (motion), and the reference also represents an orientation through angles. Hu)
Claim 11 is rejected under the same analysis as claim 4.
Claim 18 is rejected under the same analysis as claim 4.
Claims 5, 12 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Hu in view of Li and further in view of Mirhosseini et. al. (US Pub. No. 20240303766 A1 ) .
As per claim 5 Hu in view of Li teaches “The method of Claim 1, wherein applying the warping comprises, for each set of training image frames:”,(Paragraph 38 also shows a subset of frames. Hu) however Hu in view of Li does not completely teach “generating a warp field for each of a subset of the training image frames; and applying the generated warp fields to the subset of the training image frames; and wherein each warp field defines that each pixel of an image frame is warped independently of other pixels but neighboring pixels of the image frame are warped with a same or similar direction and a same or similar strength.”,
Mirhosseini teaches “generating a warp field for each of a subset of the training image frames; and applying the generated warp fields to the subset of the training image frames; (See paragraphs 90-100 and the equations“[0097] According to some implementations, a warp may be defined as a vector field V… that describes how each point in the source image 705 (IS) should be translated in order to produce the target image 707 (IW). For a particular point xS in the source image 705 (IS), the warped image coordinates xW are given by the following equation (2), which is similar to equation (1) above… In other words, for a particular pixel at point xW in the target image 707 (IW), the inverse warping operation attempts to find the location(s) xS in the source image 705 (IS) that satisfy equation (2). FPI may be used to converge to the solution in a fast and efficient manner.” ) and wherein each warp field defines that each pixel of an image frame is warped independently of other pixels but neighboring pixels of the image frame are warped with a same or similar direction and a same or similar strength.” (See paragraphs 144-154 and 171-213 “[0144] In some implementations, the neighborhood characterization vector generator 1420 is configured to generate characterization vectors 1425 for each A×B pixel neighborhood within the one or more reference image frames 842. According to some implementations, a respective characterization vector among the characterization vectors 1425 for a respective neighborhood includes a dominant movement direction for the respective neighborhood relative to the viewpoint, object motion within the respective neighborhood relative, deviation of motion for the respective neighborhood, a background depth value for the respective neighborhood, a foreground depth value for the respective neighborhood, a histogram representation of depth for the respective neighborhood, the mean depth value for the respective neighborhood, the mode value for depth in the respective neighborhood, and/or the like.” See also fig. 14 and fig. 15. See also paragraphs 162-170, pixels are warped independently. Mirhosseini)
It would have been obvious to one of ordinary skill in the art before the effective filing
date of the claimed invention to combine the teachings of Hu with the teachings of Li and Mirhosseini to generate and apply a warp field for warping pixels based on the neighbors . The modification would have been motivated by the desire to reduce consumption and converse time, to account for chromatic aberration and to improve warping by being adaptive and statistically more robust by choosing the best results, therefore it is an improvement, as suggested by Mirhosseini (See paragraphs 32-33 and 195 “[0033] Various implementations disclosed herein include devices, systems, and methods for inverse iterative warping based on an adaptive statistically robust warp (ASRW) algorithm...The method includes obtaining a reference image frame and forward flow information associated with the reference image frame; obtaining a plurality of characterization vectors for each of a plurality of neighborhoods of pixels in the reference image frame,” See also paragraphs 203-204 “[0203] In order to combat this problem, in some implementations, the methods described herein (e.g., the method 1300 associated with the SRW algorithm in FIG. 13 and the method 2000 associated with the ASRW algorithm in FIG. 20) perform multiple inverse warp operations from a plurality of starting points with varying depths for each pixel and chooses the best result from among the multiple inverse warp operations. However, in practice, the SRW or ASRW algorithms may be performed on each sub-pixel of an RGB display type in order to account for chromatic aberration that occurs therein.” Mirhosseini)
Claim 12 is rejected under the same analysis as claim 5.
Claim 19 is rejected under the same analysis as claim 5.
Claims 6, 13 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Hu in view of Li and Mirhosseini and further in view of Herbreteau et. al. (Herbreteau, Sébastien, Emmanuel Moebel, and Charles Kervrann. "Normalization-equivariant neural networks with application to image denoising." Advances in Neural Information Processing Systems 36 (2023): 5706-5728.) .
As per claim 6, Hu in view of Li and Mirhosseini already teaches “The method of Claim 5, wherein generating the warp field for each of the subset of the training image frames comprises:”, however Hu in view of Li and Mirhosseini does not teach “generating white Gaussian noise; and applying a linear two-dimensional (2D) Gaussian blur operator and normalization to the white Gaussian noise.”
Herbreteau teaches “generating white Gaussian noise; and applying a linear two-dimensional (2D) Gaussian blur operator and normalization to the white Gaussian noise.” (See page 3 section 3.2 paragraphs 1-7 and equations 1-3 “Since scaling up an image by a positive factor λ or adding it up a constant shift µ does not change its contents, it is natural to expect scale and shift equivariance, i.e. normalization equivariance, from the denoising procedure emulated by f. In image denoising, a majority of methods usually assume an additive white Gaussian noise model with variance σ^2. The corruption model then reads y ∼ N(x,σ^2 In), where In denotes the identity matrix of size n, and the noise standard deviation σ > 0 is generally passed as an additional argument to the denoiser (“non-blind” denoising)… The most rudimentary methods for image denoising are the smoothing filters, among which we can mention the averaging filter or the Gaussian filter for the linear filters and the median filter which is nonlinear. These elementary “blind” denoisers all implement a normalization-equivariant function. More generally, one can prove that a linear filter is normalization equivariant if and only if its coefficients add up to 1. In others words, normalization-equivariant linear filters process images by affine combinations of pixels… See also page 4 section 3.3 paragraphs 1-4 “Usually, the weights of neural networks are learned on a training set containing data all normalized to the same arbitrary interval [a0,b0]. This training procedure improves the performance and allows for more stable optimization of the model. At inference, unseen data are processed within the interval [a0, b0] via a a-b linear normalization with a0 ≤ a < b ≤ b0 denoted Ta,b and defined by: Ta,b : y → (b −a) y−min(y) max(y) −min(y) +a. (4) Note that this transform is actually the unique linear one with positive slope that exactly bounds the output to [a,b].” See also sections 4 including 4.1-4.3 and section 5 including 5.1-5.2. See also appendix C.2. Herbreteau )
It would have been obvious to one of ordinary skill in the art before the effective filing
date of the claimed invention to combine the teachings of Hu with the teachings of Li, Mirhosseini and Herbreteau to generate white Gaussian noise and apply a linear operator for normalization when generating the warp field. The modification would have been motivated by the desire to improve performance and more stable optimization when training, as suggested by Herbreteau (See page 4 section 3.3 “Deep learning hides a subtlety about normalization equivariance that deserves to be highlighted. Usually, the weights of neural networks are learned on a training set containing data all normalized to the same arbitrary interval [a0,b0]. This training procedure improves the performance and allows for more stable optimization of the model.” Herbretau)
Claim 13 is rejected under the same analysis as claim 6.
Claim 20 is rejected under the same analysis as claim 6.
Pertinent Prior Art
Burgert et. al. (US Pub. No. 20260134518 A1) also shows warping fields with Gaussian noise .
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DYLAN J MENDEZ MUNIZ whose telephone number is (703)756-5672. The examiner can normally be reached M-F, 8AM - 5PM ET.
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, Vu Le can be reached at (571) 272-7332. 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.
/DYLAN JOHN MENDEZ MUNIZ/Examiner, Art Unit 2675
/VU LE/Supervisory Patent Examiner, Art Unit 2668
Prosecution Insights