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 .
Notice to Applicants
This action is in response to the Application filed on 01/25/2025.
Claims 1-20 are pending.
Priority
This Application claims priority to Provisional Application 63/562,717 with filing date 03/08/2024, which is acknowledged.
Information Disclosure Statement
The Information Disclosure Statement (IDS) filed on 01/25/2025 has been fully 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 9 and 19 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.
Regarding claim 9, the claim recites “activating corresponding adapter modules in each adapter layer that are trained in the fine-tuning phase for the given degradation type”, but parent claim 1 specifically recites “each adapter layer including a pre-trained module in parallel with an adapter module” (emphasis added to each quotation). Thus, claim 9 creates confusion as to the amount of adapter modules in each adapter layer.
Regarding claim 19, this claim is rejected for the same reasons as claim 9 above.
The examiner suggests amending claims 9 and 19 to additionally recite “wherein each adapter layer includes a plurality of adapter modules” to overcome these rejections.
35 USC § 101 Analysis
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The examiner determines that all of claims 1-20 are eligible under 101. The 101 analysis is presented below for a clear record.
Analysis for claim 1 is provided in the following. Claim 1 is reproduced in the following (annotation added):
A method for image restoration, comprising:
receiving a low-quality (LQ) image having a given degradation type;
processing the LQ image by a neural network that includes a series of adapter layers, each adapter layer including a pre-trained module in parallel with an adapter module,
wherein the pre-trained module has been trained in a pre-training phase by images having a plurality of degradation types, and the adapter module has been trained in a fine-tuning phase subsequent to the pre-training phase by images having the given degradation type;
in each adapter layer, adding a first output of the pre-trained module and a second output of the adapter module to produce an output of the adapter layer;
and generating a high-quality (HQ) image restored from the LQ image based on outputs of the adapter layers.
Step 1: Does the claim belong to one of the statutory categories? Claim 1 is directed to a process, which is a statutory category of invention (YES).
Step 2A Prong One: Does the claim recite a judicial exception? Step e is regarded as reciting mathematical calculations, by reciting adding the outputs of the two modules together (YES).
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application? Step a is a non-limiting preamble. Step b recites mere data gathering. Step c, however, recites processing the LQ image through a neural network with a specific architecture, and step d recites specific training details of said architecture. Although the processing and training necessarily involve mathematical calculations to some degree, the limitations themselves are not directed to any abstract ideas. Finally, step f recites generating a HQ image from the LQ image, based on the outputs. The specific architecture, training details, and HQ image generation all integrate the mathematical calculations into a practical application (YES). Claim 1 is eligible.
Similar analysis is applicable to independent claim 11, which additionally recites a computerized system at a high level of generality. Claim 11 is eligible.
Claims 2-3 and 12-13 each recite further mathematical calculations, but these are still integrated into the same practical application. Claims 3 and 13 in particular also recite further architecture details. Claims 2-3 and 12-13 are eligible.
Claims 4-8 and 14-18 each recite additional elements of further training details with no new judicial exceptions. Claims 4-8 and 14-18 are eligible.
Claims 9 and 19 recite identifying the given degradation type of the LQ image, which is practically performable in the human mind. However, the claims further recite activating corresponding adapter modules corresponding to said identified type, which integrates this into a practical application. Claims 9 and 19 are eligible.
Claims 10 and 20 recite further architecture details with no new judicial exceptions. Claims 10 and 20 are eligible.
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.
Claims 1-2, 4-5, 10-12, 14-15, and 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Sun et al. ("Adamer: Adapting Transformer Modules For Training-Efficient Image Restoration", CAC paper, November 2023).
PNG
media_image1.png
264
1167
media_image1.png
Greyscale
Sun Figure 2
PNG
media_image2.png
349
489
media_image2.png
Greyscale
Sun Figure 2(a)
PNG
media_image3.png
301
669
media_image3.png
Greyscale
Annotated Sun Figure 2(b)
Regarding claim 1, Sun discloses a method for image restoration (see figure 2), comprising:
receiving a low-quality (LQ) image having a given degradation type (see figure 2, degraded image and section III.A, “Network Architecture”);
processing the LQ image by a neural network that includes a series of adapter layers (see figure 2, where the degraded image is input to the Adamer neural network with a plurality of ATB/adapter layers),
each adapter layer including a pre-trained module in parallel with an adapter module (see figure 2(b) and section III.B, "Adaptation Transformer Block", where each ATB includes a central pipeline/pre-trained module in the form of Norm -> MDTA -> Adapter (annotated "A1" by the examiner) -> Addition -> Norm -> MLP; figure 2(a) shows that each adapter's output is added to its input, thus adapter A1 is considered to be in parallel with the main pipeline; the end of the pipeline also includes another adapter (annotated "A2" by the examiner) that executes in a separate branch, and is thus also considered parallel to the main pipeline),
wherein the pre-trained module has been trained in a pre-training phase by images having a plurality of degradation types (page 1 of 5, right column and section V, “Conclusions” specify that the model is generally pre-trained for image restoration before addition of the adapter modules),
and the adapter module has been trained in a fine-tuning phase subsequent to the pre-training phase by images having the given degradation type (see page 3 of 5, right column, where in further training, only the adapter weights are updated; the adapters are necessarily trained on at least one degradation type, given the examples of deraining, denoising, and deblurring shown in table 1);
in each adapter layer, adding a first output of the pre-trained module and a second output of the adapter module to produce an output of the adapter layer (see figure 2(b), where the outputs of the main pipeline and adapter branch are added together as the final output of each ATB);
and generating a high-quality (HQ) image restored from the LQ image based on outputs of the adapter layers (see figure 2, where the original degraded image is finally added to the convolved output of the last ATB to generate a restored/HQ image).
Regarding claim 2, Sun discloses perform convolution operations on an output of a last adapter layer in the series of adapter layers (see figure 2 and section III.A, "Network Architecture", where a final 3x3 convolution is performed after the final ATB);
and add the LQ image to an output of the convolution operations to obtain the HQ image (see figure 2, where the original degraded image is finally added to the convolved output of the last ATB to generate a restored/HQ image).
Regarding claim 4, Sun discloses wherein parameters of the pre-training module in each adapter layer stay unchanged in the fine-tuning phase (see page 3 of 5, right column, where in further training, only the adapter weights are updated).
Regarding claim 5, Sun discloses wherein the adapter module in each adapter layer is deactivated in the pre-training phase (page 1 of 5, right column specifies that the adapters are introduced after the pre-trained model is frozen, thus the adapters are not trained during the pre-training).
Regarding claim 10, Sun discloses wherein the series of adapter layers are grouped into a plurality of adapter blocks, and the adapter modules in different adapter blocks have different feature sizes in one or more dimensions (see figure 2 and section III.A, "Network Architecture", where each ATB of a feature dimension on the encoder side can be considered grouped with the ATB of the same dimensions on the decoder side).
Regarding claim 11, Sun discloses a system for image restoration, comprising: a plurality of processors (see figure 4 and section IV.E, "Training-Efficient Result", where the model is trained on a GPU, which are known in the art to include a plurality of processing cores);
and memory to store parameters of a neural network that includes a series of adapter layers (executing any modern neural net requires a memory to store parameters of said network; page 1 of 5, left column specifies that the adapters are designed to be memory-friendly).
The remainder of claim 11 recites steps identical to those of claim 1. Therefore, Sun anticipates claim 11 as applied to claim 1 above.
Regarding claim 12, Sun discloses claim 12 as applied to claim 2 above.
Regarding claim 14, Sun discloses claim 14 as applied to claim 4 above.
Regarding claim 15, Sun discloses claim 15 as applied to claim 5 above.
Regarding claim 20, Sun discloses claim 20 as applied to claim 10 above.
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.
Claims 3 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Sun et al. ("Adamer: Adapting Transformer Modules For Training-Efficient Image Restoration", CAC paper, November 2023) in view of Chen et al. ("Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation", ECCV paper, 22 August 2018).
Regarding claim 3, Sun fails to disclose the new limitations of claim 3.
Pertaining to the same field of endeavor, Chen discloses wherein the adapter module in each adapter layer includes a first pointwise convolution layer (PConv) connected in parallel to a depth-wise convolution layer (DConv) (see figure 2 and section 3.1, “Encoder-Decoder with Atrous Convolution”, where the encoder side uses both 1x1 pointwise convolutions and 3x3 depthwise atrous convolutions),
and respective outputs of the first PConv and the DConv are added together and processed by a second PConv to produce the second output (see figure 2, where all the convolution results are then applied to a final 1x1 pointwise convolution).
Sun and Chen are considered analogous art, as they are both directed to convolutional neural nets for image processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have integrated the teachings of Chen into Sun by replacing the encoder-decoder of Sun's adapters with Chen's encoder-decoder because depthwise separable convolution significantly reduces computation while maintaining high performance (see Chen section 3.1).
Regarding claim 13, Sun in view of Chen discloses claim 13 as applied to claim 3 above.
Claims 6-9 and 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Sun et al. ("Adamer: Adapting Transformer Modules For Training-Efficient Image Restoration", CAC paper, November 2023) in view of Yang et al. ("Visual Recognition-Driven Image Restoration for Multiple Degradation with Intrinsic Semantics Recovery", IEEE/CVF CVPR paper, 22 August 2023).
Regarding claim 6, Sun fails to disclose the new limitations of claim 6.
Pertaining to the same field of endeavor, Yang discloses training multiple sets of adapter modules in the finetuning phase, each set trained to restore a corresponding degradation type in input images (see figure 2 and section 3.2, "Intrinsic Semantic Enhancement", where a plurality of Degradation Normalization and Compensation (DNC) modules are each independently trained on a specific degradation type).
Sun and Yang are considered analogous art, as they are both directed to convolutional neural nets for image restoration. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have integrated the teachings of Yang into Sun by replacing Sun's single adapters with Yang's multi-branch structure because each degradation type has different feature distributions that can be separately trained on (see Yang page 14063, left column, first paragraph).
Regarding claim 7, Sun fails to disclose the new limitations of claim 7.
Pertaining to the same field of endeavor, Yang discloses wherein each of the multiple sets of adapter modules is trained to restore one of the plurality of degradation types (see citations to claim 6 above, where each DNC is trained on a specific degradation type).
Sun and Yang are considered analogous art, as they are both directed to convolutional neural nets for image restoration. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have integrated the teachings of Yang into Sun by replacing Sun's single adapters with Yang's multi-branch structure because each degradation type has different feature distributions that can be separately trained on (see Yang page 14063, left column, first paragraph).
Regarding claim 8, Sun fails to disclose the new limitations of claim 8.
Pertaining to the same field of endeavor, Yang discloses wherein one of the multiple sets of adapter modules is trained to restore a new degradation type for which the pre-trained module in each adapter layer has not been trained (see figure 2, feature extractor and page 14061, right column, last paragraph, where the feature extractors are generally trained on large-scale datasets, as opposed to the specific degradations of the DNCs).
Sun and Yang are considered analogous art, as they are both directed to convolutional neural nets for image restoration. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have integrated the teachings of Yang into Sun by replacing Sun's single adapters with Yang's multi-branch structure because each degradation type has different feature distributions that can be separately trained on (see Yang page 14063, left column, first paragraph).
Regarding claim 9, Sun fails to disclose the new limitations of claim 9.
Pertaining to the same field of endeavor, Yang discloses identifying the given degradation type of the LQ image; and activating corresponding adapter modules in each adapter layer that are trained in the fine-tuning phase for the given degradation type (see figure 2 and section 3.2, "Intrinsic Semantic Enhancement", where the input image features are sent to each trained DNC module to process degradation-specific features, which then guide the FGM for enhancement; the DNC corresponding to the degradation type will naturally be most influential in said guiding/identifying).
Sun and Yang are considered analogous art, as they are both directed to convolutional neural nets for image restoration. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have integrated the teachings of Yang into Sun by replacing Sun's single adapters with Yang's multi-branch structure because each degradation type has different feature distributions that can be separately trained on (see Yang page 14063, left column, first paragraph).
Regarding claim 16, Sun in view of Yang discloses claim 16 as applied to claim 6 above.
Regarding claim 17, Sun in view of Yang discloses claim 17 as applied to claim 7 above.
Regarding claim 18, Sun in view of Yang discloses claim 18 as applied to claim 8 above.
Regarding claim 19, Sun in view of Yang discloses claim 19 as applied to claim 9 above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS JOHN HELCO whose telephone number is (703)756-5539. The examiner can normally be reached on Monday-Friday from 9:00 AM to 5:00 PM.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Matthew Bella, can be reached at telephone number 571-272-7778. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center for authorized users only. Should you have questions about access to Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form.
/NICHOLAS JOHN HELCO/Examiner, Art Unit 2667 /MATTHEW C BELLA/Supervisory Patent Examiner, Art Unit 2667
Prosecution Insights