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 .
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 1-18 are rejected under 35 U.S.C. 103 as being unpatentable over Kabaria et al. (US Pub No. 20210304364 A1) in view of Schmidt et al. (US Pub No. 20210241474 A1).
Regarding Claim 1,
Kabaria discloses A device comprising: and one or more processors configured to: (Kabaria, [0059], discloses a system for practicing those methods. The disclosed methods may be performed by a combination of hardware, software, firmware, middleware, and computer-readable medium (collectively computer) installed in and/or communicatively connected to a user device. Systems, apparatuses, and methods described herein may be implemented using digital circuitry, or using one or more computers using well-known computer processors, memory units, storage devices, computer software, and other components. Typically, a computer includes a processor for executing instructions and one or more memories for storing instructions and data. A computer may also include, or be coupled to, one or more mass storage devices, such as one or more magnetic disks, internal hard disks and removable disks, magneto-optical disks, optical disks; device system with memory to store image and other data is disclosed)
apply synthetic noise to the input image to generate a noise-added image; (Kabaria, [0033], Fig. 4, discloses a machine learning workflow 400 according to an embodiment in which a denoiser 420 and a discriminator 425 are each implemented as machine learning models. In this example, workflow 400 represents an operation before any training is performed for denoiser 420 and discriminator 425. As shown, original document image 401 is provided as an input to a random noise generator 402 that adds one or more forms of document noise to generate noisy document image 403; random (synthetic) noise is added to the document input image) and
apply a denoiser to the noise-added image to generate an output image that has less noise than the input image. (Kabaria, [0033-0034], discloses original document image 401 is provided as an input to a random noise generator 402 that adds one or more forms of document noise to generate noisy document image 403, which is then input to denoiser 420. Denoiser 420 generates denoised output document image 421, which is provided to discriminator 425 for inspection. In this workflow, discriminator 425 determines (e.g., predicts) the authenticity of output document image 421, e.g., predicts whether it is an original document image or not. In this example (e.g., before any training is performed on denoiser 420), output document image 421 contains enough noise (see shading) such that discriminator 425 determines/predicts that it is not an original document image. Feedback 430 is then provided from discriminator 425 to denoiser 420 to train denoiser 420; the process continues iteratively, such that denoiser 420 learns to adapt and generate output document images 421 that more closely resemble original images (e.g., improves the denoising functionality), while discriminator 425 learns features of original and denoised document images (e.g., by learning aggregate underlying patterns that constitute noise and patterns that make a document image an original document image), thereby improving the predictive functionality. Accordingly, workflow 400 is meant to illustrate an unconstrained mode of operation because learning is not constrained to side-by-side comparative analysis, but instead is based on learning underlying patterns that can be more broadly adapted and applied; denoiser is applied to the random noise generated input document image to denoise and the output generated image is not same as original image with improved quality (less noise) than the input image).
Kabaria does not explicitly disclose a memory configured to store an input image;
Schmidt discloses a memory configured to store an input image; (Scmidt, [0023], discloses computer system 130 can either generate images or retrieve previously stored computer graphic images such as frame 124. Since the CG images are created based on computer models, all of the depth information is already defined for each of their elements. The remaining steps of FIG. 1 are needed to quickly and accurately determine depth information for elements in the picture camera image in order that the live action image can be accurately placed “into” (i.e., composited with) the CG image; image is stored)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to combine the teachings of Kabaria in view of Schmidt having a method to add synthetic noise and further denoise the image to remove noise from the input image, with the teachings of Schmidt storing input image to memory for storing the image and accessing it as required for processing the image for denoising and outputting denoised image.
Regarding Claim 2,
The combination of Kabaria and Schmidt further discloses wherein the denoiser is a blind denoiser. (Kabaria, [0031], discloses the discriminator eventually learns what makes an image original or unoriginal. The differentiating factor is, in the aggregate, noise, so the discriminator learns to identify noise in the image through the training sequences. Initially, both models, the denoiser and discriminator (sometimes referenced as the detective) are untrained. For example, the discriminator does not initially understand what comprises original images. Over time, the discriminator learns what features are in an original image and that knowledge is backpropagated to the denoiser; the denoiser is untrained about the noise intensity present in the original image).
Regarding Claim 3,
The combination of Kabaria and Schmidt further discloses wherein the input image includes a first amount of noise, and wherein the output image includes a second amount of noise that is less than the first amount. (Kabaria, [0033-0034], discloses original document image 401 is provided as an input to a random noise generator 402 that adds one or more forms of document noise to generate noisy document image 403, which is then input to denoiser 420. Denoiser 420 generates denoised output document image 421, which is provided to discriminator 425 for inspection. In this workflow, discriminator 425 determines (e.g., predicts) the authenticity of output document image 421, e.g., predicts whether it is an original document image or not. In this example (e.g., before any training is performed on denoiser 420), output document image 421 contains enough noise (see shading) such that discriminator 425 determines/predicts that it is not an original document image. Feedback 430 is then provided from discriminator 425 to denoiser 420 to train denoiser 420; the process continues iteratively, such that denoiser 420 learns to adapt and generate output document images 421 that more closely resemble original images (e.g., improves the denoising functionality), while discriminator 425 learns features of original and denoised document images (e.g., by learning aggregate underlying patterns that constitute noise and patterns that make a document image an original document image), thereby improving the predictive functionality. Accordingly, workflow 400 is meant to illustrate an unconstrained mode of operation because learning is not constrained to side-by-side comparative analysis, but instead is based on learning underlying patterns that can be more broadly adapted and applied; denoiser is applied to the random noise generated input document image to denoise and the output generated image is not same as original image with improved quality (less noise) than the input image).
Regarding Claim 4,
The combination of Kabaria and Schmidt further discloses wherein the one or more processors are further configured to generate the synthetic noise. (Kabaria, [0030], [0063], discloses denoising system according to one or more embodiments applies the concept of generator and discriminator functions from the GAN-based neural network model. For example, a first neural network model is used as a denoiser to carry out the generator function. More specifically, the denoiser receives noisy images (e.g., clean original images from a synthetic dataset that have noise randomly added) and generates denoised output images. A second neural network model used as a discriminator is given both original images and denoised images (e.g., noisy versions of the original images that have been passed through the denoiser) to predict whether the received images are an original image or not. For example, the discriminator may output a probability that a given image is an original image. In some embodiments, the discriminator may be given equal quantities of original and denoised images, but the images are not necessarily paired and provided to the discriminator at the same time, e.g., the discriminator is not given both an original image and its corresponding noisy version that has been denoised image at the same time. In this manner, the training model is not constrained.A high-level block diagram of an exemplary computing system 1300 that may be used to implement systems, apparatus, and methods described herein is depicted in FIG. 13. In some embodiments, computing system 1300 may be one or more of the computing systems depicted and/or described herein. Computing system 1300 includes a bus 1305 or other communication mechanism for communicating information, and processor(s) 1310 coupled to bus 1305 for processing information. Processor(s) 1310 may be any type of general or specific purpose processor, including a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Graphics Processing Unit (GPU), multiple instances thereof, and/or any combination thereof. Processor(s) 1310 may also have multiple processing cores, and at least some of the cores may be configured to perform specific functions. Multi-parallel processing may be used in some embodiments; computing system 1300 further includes a memory 1315 for storing information and instructions to be executed by processor(s) 1310. Memory 1315 can be comprised of any combination of Random Access Memory (RAM), Read Only Memory (ROM), flash memory, cache, static storage such as a magnetic or optical disk, or any other types of non-transitory computer-readable media or combinations thereof. Non-transitory computer-readable media may be any available media that can be accessed by processor(s) 1310 and may include volatile media, non-volatile media, or both. The media may also be removable, non-removable, or both; processors are integrated to execute functions including generating synthetic noised image data to be added to original input document image).
Regarding Claim 5,
The combination of Kabaria and Schmidt further discloses wherein the synthetic noise is generated based on a first distribution associated with training of the denoiser. (Kabaria, [0030], [0063], discloses denoising system according to one or more embodiments applies the concept of generator and discriminator functions from the GAN-based neural network model. For example, a first neural network model is used as a denoiser to carry out the generator function. More specifically, the denoiser receives noisy images (e.g., clean original images from a synthetic dataset that have noise randomly added) and generates denoised output images. A second neural network model used as a discriminator is given both original images and denoised images (e.g., noisy versions of the original images that have been passed through the denoiser) to predict whether the received images are an original image or not. For example, the discriminator may output a probability that a given image is an original image. In some embodiments, the discriminator may be given equal quantities of original and denoised images, but the images are not necessarily paired and provided to the discriminator at the same time, e.g., the discriminator is not given both an original image and its corresponding noisy version that has been denoised image at the same time. In this manner, the training model is not constrained.A high-level block diagram of an exemplary computing system 1300 that may be used to implement systems, apparatus, and methods described herein is depicted in FIG. 13. In some embodiments, computing system 1300 may be one or more of the computing systems depicted and/or described herein. Computing system 1300 includes a bus 1305 or other communication mechanism for communicating information, and processor(s) 1310 coupled to bus 1305 for processing information. Processor(s) 1310 may be any type of general or specific purpose processor, including a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Graphics Processing Unit (GPU), multiple instances thereof, and/or any combination thereof. Processor(s) 1310 may also have multiple processing cores, and at least some of the cores may be configured to perform specific functions. Multi-parallel processing may be used in some embodiments; computing system 1300 further includes a memory 1315 for storing information and instructions to be executed by processor(s) 1310. Memory 1315 can be comprised of any combination of Random Access Memory (RAM), Read Only Memory (ROM), flash memory, cache, static storage such as a magnetic or optical disk, or any other types of non-transitory computer-readable media or combinations thereof. Non-transitory computer-readable media may be any available media that can be accessed by processor(s) 1310 and may include volatile media, non-volatile media, or both. The media may also be removable, non-removable, or both; processors are integrated to execute functions including generating synthetic noised image data to be added to original input document image and synthetic noise is generated based in iteration of feedback and training added in the original image).
Regarding Claim 6,
The combination of Kabaria and Schmidt further discloses wherein noise in the input image is associated with a second distribution that is different from the first distribution and wherein the denoiser, during removal of the synthetic noise from the noise-added image, also removes at least some of the noise of the input image. (Kabaria, [0033-0034], discloses original document image 401 is provided as an input to a random noise generator 402 that adds one or more forms of document noise to generate noisy document image 403, which is then input to denoiser 420. Denoiser 420 generates denoised output document image 421, which is provided to discriminator 425 for inspection. In this workflow, discriminator 425 determines (e.g., predicts) the authenticity of output document image 421, e.g., predicts whether it is an original document image or not. In this example (e.g., before any training is performed on denoiser 420), output document image 421 contains enough noise (see shading) such that discriminator 425 determines/predicts that it is not an original document image. Feedback 430 is then provided from discriminator 425 to denoiser 420 to train denoiser 420; the process continues iteratively, such that denoiser 420 learns to adapt and generate output document images 421 that more closely resemble original images (e.g., improves the denoising functionality), while discriminator 425 learns features of original and denoised document images (e.g., by learning aggregate underlying patterns that constitute noise and patterns that make a document image an original document image), thereby improving the predictive functionality. Accordingly, workflow 400 is meant to illustrate an unconstrained mode of operation because learning is not constrained to side-by-side comparative analysis, but instead is based on learning underlying patterns that can be more broadly adapted and applied; denoiser is applied to the random noise generated input document image to denoise and the output generated image is not same as original image with improved quality (less noise) than the input image). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 7,
The combination of Kabaria and Schmidt further discloses apply multiple versions of synthetic noise to the input image to generate multiple noise-added images; denoise each of the multiple noise-added images to generate multiple output images; and combine the multiple output images to generate an ensemble output image. (Schmidt, [0034], discloses sequence of steps in FIG. 2 for pre-processing to generate an improved disparity map can also be used to improve disparity map with artifacts 150 of FIG. 1. The picture image can be combined with disparity map with artifacts 150. In other words, each of steps 250-270 may be applied to an initial disparity map with artifacts such as 150 of FIG. 1 to generate an improved disparity map without artifacts; artifacts (noise added) images are combined in one picture frame (multiple noise added images) and output the denoised image). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 8,
The combination of Kabaria and Schmidt further discloses a display device configured to display the output image. (Kabaria, [0066-0067], discloses processor(s) 1310 are further coupled via bus 1305 to a display 1325 that is suitable for displaying information to a user. Display 1325 may also be configured as a touch display and/or any suitable haptic I/O device; keyboard 1330 and a cursor control device 1335, such as a computer mouse, a touchpad, etc., are further coupled to bus 1305 to enable a user to interface with computing system. However, in certain embodiments, a physical keyboard and mouse may not be present, and the user may interact with the device solely through display 1325 and/or a touchpad (not shown). Any type and combination of input devices may be used as a matter of design choice. In certain embodiments, no physical input device and/or display is present. For instance, the user may interact with computing system 1300 remotely via another computing system in communication therewith, or computing system 1300 may operate autonomously; output images are displayed on display screen). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 9,
Kabaria further discloses an image sensor configured to generate image data corresponding to the input image. (Kabaria, [0044], Fig. 7A, discloses process 700 starts at step 701 in which the denoiser element is initialized followed by step 702 in which the discriminator element is initialized. Dataset generation occurs in block 703 to generate images for subsequent training use. In particular, original document images are generated in step 704 to create original image dataset 705. In step 706, original images from dataset 705 are randomly sampled to produce target images 710. Noise is then added to target images 710 at step 715 to produce noisy images 716; image data is generated from the original input image).
Regarding Claim 10,
The combination of Kabaria and Schmidt further discloses a modem coupled to the one or more processors, the modem configured to receive the input image from a second device. (Schmidt, [0037], Network interface 950 typically includes an Ethernet card, a modem (telephone, satellite, cable, ISDN), (asynchronous) digital subscriber line (DSL) unit, and the like. Further, network interface 950 may be physically integrated on the motherboard of computer 920, may be a software program, such as soft DSL, or the like; modem integration is disclosed). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 11,
The combination of Kabaria and Schmidt further discloses wherein the one or more processors are integrated in a headset device that includes a display, and wherein the headset device is configured, when worn by a user, to display the output image at the display. (Schmidt, [0041] In addition to generating recorded and synthetic datasets from the actual movie set on which the filming is to take place, generic datasets may be obtained of unrelated sets or environments. Any one or more of these types of data, or mixtures or combinations of data; can be combined into a “training dataset,” used to improve the later real-time depth detection during a live-action shoot so that digital images can be more accurately composited onto, e.g., a director's camera viewfinder or an actor's virtual or augmented reality headset; in order to show what the final, composited, scene will look like; wearable headset device to display the output is disclosed).
Regarding Claim 12,
The combination of Kabaria and Schmidt further discloses wherein the one or more processors are integrated in at least one of a mobile phone, a tablet computer device, a wearable electronic device, or a camera device. (Schmidt, [0041] In addition to generating recorded and synthetic datasets from the actual movie set on which the filming is to take place, generic datasets may be obtained of unrelated sets or environments. Any one or more of these types of data, or mixtures or combinations of data; can be combined into a “training dataset,” used to improve the later real-time depth detection during a live-action shoot so that digital images can be more accurately composited onto, e.g., a director's camera viewfinder or an actor's virtual or augmented reality headset; in order to show what the final, composited, scene will look like; wearable headset device to display the output is disclosed). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 13,
The combination of Kabaria and Schmidt further discloses wherein the one or more processors are integrated in a vehicle, the vehicle further including a display device configured to display the output image. (Kabaria, [0020], [0041], discloses the live action camera rig is used to record live action such as moving actors, vehicles or other objects.addition to generating recorded and synthetic datasets from the actual movie set on which the filming is to take place, generic datasets may be obtained of unrelated sets or environments. Any one or more of these types of data, or mixtures or combinations of data; can be combined into a “training dataset,” used to improve the later real-time depth detection during a live-action shoot so that digital images can be more accurately composited onto, e.g., a director's camera viewfinder or an actor's virtual or augmented reality headset; in order to show what the final, composited, scene will look like; movie set with moving vehicle camera display is integrated with moving vehicle to capture the scenes and display as augment reality on the display screen). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Regarding Claim 14,
The combination of Kabaria and Schmidt further discloses wherein the one or more processors are included in an integrated circuit. (Kabaria, [0063], high-level block diagram of an exemplary computing system 1300 that may be used to implement systems, apparatus, and methods described herein is depicted in FIG. 13. In some embodiments, computing system 1300 may be one or more of the computing systems depicted and/or described herein. Computing system 1300 includes a bus 1305 or other communication mechanism for communicating information, and processor(s) 1310 coupled to bus 1305 for processing information. Processor(s) 1310 may be any type of general or specific purpose processor, including a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Graphics Processing Unit (GPU), multiple instances thereof, and/or any combination thereof. Processor(s) 1310 may also have multiple processing cores, and at least some of the cores may be configured to perform specific functions. Multi-parallel processing may be used in some embodiments; integrated circuit is disclosed). Additionally, the combination of Kabaria and Schmidt as applied in rejection of claim 1 apply to this claim.
Claims 15-18 recite method with steps corresponding to the device elements recited in Claims 1, 2 5 and 7 respectively. Therefore, the recited steps of the method Claims 15-18 are mapped to the proposed combination in the same manner as the corresponding elements of Claims 1, 2 5 and 7 respectively. Additionally, the rationale and motivation to combine the Kabaria and Schmidt references presented in rejection of Claim 1, apply to these claims.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 19-20 are rejected under 35 U.S.C. 102 (a)(1)/(a)(2) as being anticipated by Yang et al. (US Pub No. 20210287342 A1).
Regarding Claim 19,
Yang discloses A device comprising: a memory configured to store an input image having a first size; (Yang, [0039], discloses performing image processing with, for example, user equipment (UE) such as a standalone digital camera or a digital camera integrated into a smartphone. FIG. 1 is a block diagram of an example of a digital camera system 100 in accordance with some embodiments of the present disclosure, which may be components of, for example, a standalone digital camera or a smartphone. For the sake of clarity, a digital camera system 100 generally includes a digital camera module 110 including a lens 112 mounted in front of an image sensor 114 (e.g., a complementary metal oxide semiconductor (CMOS) image sensor). The digital camera system 100 may further include a processing circuit such as a processor (e.g., an application processor (AP) and/or an image signal processor (ISP)) 130 configured to receive data captured by the digital camera module 110 (e.g., image data of a scene), and may store the received data in memory 150. The memory 150 may include dynamic memory (DRAM) and/or persistent memory (e.g., flash memory). In some circumstances, the image signal processor 116 is integrated into the processor 130. In some embodiments, the digital camera system 100 further includes a co-processing circuit or a co-processor 170 such as a field programmable gate array (FPGA), a graphical processing unit (GPU), a vector processor, or a neural processing unit. In some embodiments, the co-processor is integrated with the processor 130 (e.g., on the same physical die). The processor and the co-processor may be referred to herein jointly as a processor or a processing circuit, as various operations may be distributed between different physical circuitry in accordance with various design choices and suitability to different types of tasks; input image data is stored in memory) and
one or more processors configured to: upsample the input image to generate an upsampled image that has a second size larger than the first size; (Yang, [0073-0075] , iThe encoded features 584 at the fourth scale 570 are supplied to an upsampling module 575 that upsamples the encoded features 584 from the fourth scale 570 to the third scale 550 to generate upsampled features 565 at the third scale 550; upsampled features 565 are concatenated with the encoded features 564 at the third scale 550 and the concatenated features are supplied to a third MRDB 556 of the third scale 550 to generate output features 566 at the third scale 550. The output features 566 at the third scale 550 are supplied to an upsampling module 553 to up sample the output features 556 from the third scale 550 to the second scale 530 and to apply a 1×1 convolution to the up sampled features to generate up sampled features 545 at the second scale 530; up sampled features 545 are concatenated with the encoded features 544 at the second scale 530 and the concatenated features are supplied to a third MRDB 536 of the second scale 530 to generate output features 546 at the second scale 530. The output features 546 at the second scale 530 are supplied to an upsampling module 531 to up sample the output features 546 from the second scale 530 to the first scale 510 and to apply a 1×1 convolution to the up sampled features to generate up sampled features 525 at the first scale 510; image feature data is up sampled to create second size image)
apply a synthetic blurring kernel to the up sampled image to generate a blurred image; (Yang, [0088], discloses neural networks for performing image processing operations using neural network architectures in accordance with embodiments of the present disclosure are trained using training data sets, which include noisy input images and corresponding denoised ground truth images (e.g., the desired non-noisy output of the network). Training a neural network generally involves initializing the neural network (e.g., setting the weights in the network, such as the weights in the convolutional kernels, to random values), and supplying the training input data to the network. The output of the network is then compared against the labeled training data to generate an error signal (e.g., a difference between the current output and the ground truth output), and a backpropagation algorithm is used with gradient descent to update the weights, over many iterations, such that the network computes a result closer to the desired ground truth image; upsampled image is applied with noise kernel to blur the image)
downsample the blurred image to generate a downsampled image; (Yang, [0089], discloses image datasets for training a convolutional neural network to perform denoising can be divided into two categories: synthetic image datasets and real image datasets based on the source of the provided noisy images within dataset. Synthetic image datasets are usually built by: first collecting high-quality images as noise-free images by downsampling a high-resolution image or post-processing a low-ISO image; then adding synthetic noise based on statistic noise models (e.g., a Gaussian noise model or a Poissonian-Gaussian noise model) to generate synthetic noisy images. Real image datasets are generally generated by: First collecting multiple real noisy images in a short time (e.g., to ensure minimal image content change, such as scene luminance change or movement of objects in a scene; then fusing these multiple images to generate a synthetic noise-free or low-noise image; blurred (noisy) image is downsampled) and
process the downsampled image using a super-resolution model to generate an output image. (Yang, [0047] An MRDB according to embodiments of the present disclosure combines multi- scale features computed by an atrous spatial pyramid pooling (ASPP) module (see, e.g., L. C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In ECCV, 801-818, 2018.) and other features computed by a residual dense block (RDB) module (see, e.g., Y. Zhang, Y. Tian, Y. Kong, B. Zhong and Y. Fu. Residual dense network for image super-resolution. In CVPR, pp. 2472-2481, 2018.), where the input feature map 302 is supplied to ASPP module 320 and the RDB module 350; super resolution imatge is output) .
Regarding Claim 20,
wherein the synthetic blurring kernel matches a blurring kernel used to generate low-resolution images during training of the super-resolution model. (Yang, [0036] , discloses image denoising reduces or removes the presence of noise, reconstructs details in the structural content of images, and generates higher-quality output images from lower-quality input images. Some techniques for image denoising generally relates to removing noise from RGB data (e.g., sRGB data). These include classical methods using handcrafted or explicitly specified filters, such as local mean and block-matching and 3D filtering (BM3D). In addition, a neural network architecture such as a convolutional neural network (CNN) provides machine learning-based alternatives to the comparative handcrafted techniques, where statistical models are automatically trained to denoise images based on large sets of training data (e.g., sets of noisy images and corresponding low-noise versions); low resolution images are generated during training).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
US 20150296152 A1 (Filtering sensor data is described, for example, where filters conditioned on a local appearance of the signal are predicted by a machine learning system, and used to filter the sensor data. In various examples the sensor data is a stream of noisy video image data and the filtering process denoises the video stream. In various examples the sensor data is a depth image and the filtering process refines the depth image which may then be used for gesture recognition or other purposes. In various examples the sensor data is one dimensional measurement data from an electric motor and the filtering process denoises the measurements. In examples the machine learning system comprises a random decision forest where trees of the forest store filters at their leaves. In examples, the random decision forest is trained using a training objective with a data dependent regularization term)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PINALBEN V PATEL whose telephone number is (571)270-5872. The examiner can normally be reached M-F: 10am - 8pm.
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.
/Pinalben Patel/Examiner, Art Unit 2673