1DETAILED 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 .
Information Disclosure Statement
The information disclosure statement (IDS) was submitted on September 23, 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 22-28 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 “JPEG-like” in claims 22-28 is a relative term which renders the claim indefinite. The term “JPEG-like” 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.
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 (i.e., changing from AIA to pre-AIA ) 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, 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, 2, 4, 9-13, 15, 16, 22 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Deetlefs US 20260030547 A1 (hereinafter referred to as “Deetlefs”) in further view of Lefebvre et al., US 20260136032 A1 (hereinafter referred to as “Lefebvre”).
Regarding claim 1, Deetlefs discloses a method for an asymmetrically-powered neural image compression with shallow decoders (Deetlefs: “(encoder 406) and the decoder 410 can have an asymmetric architecture,” and “[u]sing a more lightweight decoder 410,” (i.e., shallow decoder). ¶ [0112]) the method comprising:
determining an asymmetric computation budget between an encoding process and a decoding process, wherein the asymmetric computation budget comprises an encoding computation budget that is greater than a decoding computation budget, performing the encoding process using an encoder (Deetlefs: “…an asymmetrical architecture, where the encoder has a higher complexity (e.g. is larger) than the decoder.” ¶ [0068]. “The autoencoder formed of the first machine learning model (encoder 406) and the decoder 410 can have an asymmetric architecture, wherein first machine learning model 406 can be larger and more complex than the decoder 410…[u]sing a more lightweight decoder 410 allows more computational resource to be dedicated to resource-intensive tasks performed by the first machine learning model 406” ¶ [0122]),
Deetlefs does not explicitly disclose wherein the encoding process comprises generating encoded data from a source data using an analysis transform; and performing the decoding process using a shallow decoder, wherein the decoding process comprises generating decoded data from the encoded data using a synthesis transform generates reconstructed input.
More specifically, Deetlefs does not explicitly disclose an encoding using an analysis transform and decoding using a synthesis transform, as recited in the claim.
However, in the same field of endeavor, Lefebvre discloses the encoding process comprises generating encoded data from a source data using an analysis transform (Lefebvre: “Lightweight decoder” (i.e., shallow decoder). ¶ ¶ [0053] and [0098]. “[I]nput X is fed into the encoder neural network… (210, also known as analysis transform).” ¶ [0041]); and performing the decoding process using a shallow decoder (Lefebvre: “lightweight decoder” (i.e., shallow decoder). ¶ ¶ [0053] and [0098]), wherein the decoding process comprises generating decoded data from the encoded data using a synthesis transform generates reconstructed input (Lefebvre: “The decoder network… (240, also known as synthesis transform generates.” ¶ [0043]. Encoder 210, decoder 240. Fig. 2).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with encoding using an analysis transform and decoding using a synthesis transform, as taught by Lefebvre, in order to improve coding efficiency.
Regarding claim 2, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the encoding and decoding computation budgets are measured in computational complexity (Deetlefs: “…an asymmetrical architecture, where the encoder has a higher complexity (e.g. is larger) than the decoder.” ¶ [0068]. “The autoencoder formed of the first machine learning model (encoder 406) and the decoder 410 can have an asymmetric architecture, wherein first machine learning model 406 can be larger and more complex than the decoder 410…[u]sing a more lightweight decoder 410 allows more computational resources to be dedicated to resource-intensive tasks performed by the first machine learning model 406” ¶ [0122]).
Regarding claim 4, the combination of Deetlefs and Lefebvre disclose the method of Claim 2, wherein the computational complexity is measured by a number of floating point operations ("FLOPs").
Deetlefs does not explicitly disclose the computational complexity is measured by a number of floating point operations ("FLOPs").
However, Lefebvre discloses computational complexity is measured by a number of floating point operations FLOPS (Lefebvre: “[F]loating-point operations have a high computational complexity and memory management which may be an obstacle for a large deployment in low-end devices.” ¶ [0054]- [0055]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with measuring computational complexity by a number of floating point operations FLOPs, as taught by Lefebvre, to determine computational complexity in order to improve coding efficiency.
Regarding claim 9, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the encoding process is selected based on the encoding computation budget and a predetermined rate-distortion ("R-D") performance.
Deetlefs does not disclose encoding process is selected based on the encoding computation budget and a predetermined rate-distortion ("R-D") performance.
However, in the same field of endeavor, Lefebvre discloses encoding process is selected based on the encoding computation budget and a predetermined rate-distortion ("R-D") performance (Lefebvre: “[T]he encoder and decoder neural networks are learned during training by minimizing a loss function defined as R+λ.Math.D, where R is a rate and D a distortion between an input image x and a reconstructed image {circumflex over (x)}. The rate R is for example defined as −log(p.sub.ψ(ŷ) and the distortion D as |x−{circumflex over (x)}|.sub.2.sup.2, where x is the input image, {circumflex over (x)} is the reconstructed image, y is the latent representation (i.e. the output of the encoder), ŷ is the quantized latent and ψ represent the parameters of the cumulative distribution functions (CDFs). “¶ [0052]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with encoding process is selected based on the encoding computation budget and a predetermined rate-distortion ("R-D") performance, as taught by Lefebvre, in order to improve coding efficiency.
Regarding claim 10, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the decoding process is selected based on the decoding computation budget and the predetermined R-D performance.
Deetlefs does not disclose wherein the decoding process is selected based on the decoding computation budget and the predetermined R-D performance.
However, in the same field of endeavor, Lefebvre discloses wherein the decoding process is selected based on the decoding computation budget and the predetermined R-D performance (Lefebvre: “[T]he encoder and decoder neural networks are learned during training by minimizing a loss function defined as R+λ.Math.D, where R is a rate and D a distortion between an input image x and a reconstructed image {circumflex over (x)}. The rate R is for example defined as −log(p.sub.ψ(ŷ) and the distortion D as |x− {circumflex over (x)}|.sub.2.sup.2, where x is the input image, {circumflex over (x)} is the reconstructed image, y is the latent representation (i.e. the output of the encoder), ŷ is the quantized latent and ψ represent the parameters of the cumulative distribution functions (CDFs).” ¶ [0052]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with decoding process is selected based on the decoding computation budget and the predetermined R-D performance, as taught by Lefebvre, in order to improve coding efficiency.
Regarding claim 11, Deetlefs discloses the method of Claim 1, wherein the encoding process comprises generating encoded data from the source data using an iterative optimization process (Deetlefs: “[I]n the self-supervised learning process, a masked autoencoder may be trained with the first training dataset. The encoder portion of the masked autoencoder may then be used as the first machine learning model. The self-supervised learning process may be an iterative process where the machine learning model is iteratively adjusted to reduce (minimise) a difference between data reconstructed by the first machine learning model and corresponding data in the first training dataset.” ¶ [0062] – [0063]).
Regarding claim 12, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the analysis transform is a neural image compression analysis transform.
Deetlefs does not explicitly disclose wherein the analysis transform is a neural image compression analysis transform.
More specifically, Deetlefs discloses a neural image analysis (Deetlefs: encoder and decoder neural networks. Abstract) but does not explicitly disclose the analysis transform.
However, in the same field of endeavor, Lefebvre discloses the analysis transform is a neural image compression analysis transform (Lefebvre: “Lightweight decoder” (i.e., shallow decoder). ¶ ¶ [0053] and [0098]. “[I]nput X is fed into the encoder neural network… (210, also known as analysis transform).” ¶ [0041]).
Since, Lefebvre teaches that an encoder neural network is also known as an analysis transform. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs such that the analysis transform is a neural image compression analysis transform., as taught Lefebvre, in order to improve coding efficiency.
Regarding claim 13, Deetlefs discloses the method of Claim 12, wherein the encoding process further comprises iterative encoding (Deetlefs: “[T]he process of FIG. 4A may be iteratively repeated.” ¶ [0123]. “FIG. 4A, the masked data element 402 is provided as an input to the first machine learning model 406, which is configured as an encoder.” ¶ [0118]).
Regarding claim 15, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the synthesis transform is a two-layer nonlinear synthesis transform.
Deetlefs does not explicitly disclose the synthesis transform is a two-layer nonlinear synthesis transform.
However, Lefebvre discloses a synthesis transform is a two-layer nonlinear synthesis transform (Lefebvre: “The encoder and decoder encoder neural networks are composed of multiple layers, such as convolutional layers. Each layer can be described as a function that first multiplies the input by a weight, adds a vector called the biases and then applies a nonlinear function (an activation function) on the resulting values.” ¶ [0044. Figs. 3, 4A, and 4B]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with a two-layer nonlinear synthesis transform, as taught by Lefebvre, in order to improve decoding performance. Lefebvre, [0053].
Regarding claim 16, Deetlefs does not explicitly disclose the method of Claim 15, wherein the two-layer nonlinear synthesis transform comprises: a first transposed convolution layer having a convolution kernel size parameter (ki), an upsampling stride parameter (si), and N output channels; an optional residential connection comprising a residual convolution transposed layer; and a second transposed convolution layer having a convolution kernel size parameter (k2), an upsampling stride parameter (s2), and a number of output channels specified by that of an image being reconstructed.
However, in the same field of endeavor, Lefebvre discloses the two-layer nonlinear synthesis transform (Lefebvre: “Decoder/Synthesis” 320. Fig. 3) comprises: a first transposed convolution layer having a convolution kernel size parameter (ki), an upsampling stride parameter (si), and N output channels; an optional residential connection comprising a residual convolution transposed layer; and a second transposed convolution layer having a convolution kernel size parameter (k2), an upsampling stride parameter (s2), and a number of output channels specified by that of an image being reconstructed (Lefebvre: “Decoder/Synthesis” 320 Fig. 3. “[Decoder/Synthesis] (320) comprises 4 deconvolutional layers Decony and 3 nonlinear inverse Generalized Divisive Normalizations (iGDN). The parameters of the convolutional layers are denoted as number of filters×kernel support height×kernel support/down- or upsampling stride, e.g., as M×5×5. In an example, M=192 or 128 and stride is equal to 2. Downsampling is applied on the encoder side and upsampling on the decoder side.” ¶ [0051]. The residential connection has been interpreted as being an optional element of the claimed two-layer synthesis transform).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with the two-layer nonlinear synthesis transform comprising: a first transposed convolution layer having a convolution kernel size parameter (ki), an upsampling stride parameter (si), and N output channels; an optional residential connection comprising a residual convolution transposed layer; and a second transposed convolution layer having a convolution kernel size parameter (k2), an upsampling stride parameter (s2), and a number of output channels specified by that of an image being reconstructed, as taught by Lefebvre, in order to reduce decoding complexity.
Regarding claim 22, the combination of Deetlefs and Lefebvre disclose the method of Claim 1, wherein the synthesis transform is a JPEG-like synthesis transform.
Deetlefs does not disclose the synthesis transform is a JPEG-like synthesis transform.
However, in the same field of endeavor Lefebvre discloses the synthesis transform is a JPEG-like synthesis transform (Lefebvre: Decoder/Synthesis 320 Figs. 3, 4A and 4B. Disclosure relates to ISO/MPEG and ITU and video coding standard H.266/VVC coding standards. ¶ [0004]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs such that the discloses the synthesis transform is a JPEG-like synthesis transform, as taught by Lefebvre, in order to improve coding efficiency.
Regarding claim 23, Deetlefs does not explicitly disclose the method of Claim 22, wherein the encoded data comprises a latent tensor of coefficients as in nonlinear transform coding, and the JPEG-like synthesis transform interprets the encoded data as coefficients of a linear synthesis transform and outputs decoded data in blocks.
However, Lefebvre discloses the encoded data comprises a latent tensor of coefficients as in nonlinear transform coding (Lefebvre: “The output of the analysis transform is Z=g.sub.a(X) that is a 3-dimensional tensor (referred to as a tensor), also called latent tensor or latent representation.” ¶ [0042]. “The GDN comprises linear transformations.” ¶ [0065], Fig. 4A), and the JPEG-like synthesis transform interprets the encoded data as coefficients of a linear synthesis transform and outputs decoded data in blocks (Lefebvre: Input Z. Fig. 2, ¶ [0043]. “Rectified Linear Unit” Fig. 4A).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs with the encoded data comprises a latent tensor of coefficients as in nonlinear transform coding, and the JPEG-like synthesis transform interprets the encoded data as coefficients of a linear synthesis transform and outputs decoded data in blocks, as taught by Lefebvre, to translates a low-dimensional, compressed representation back into high-dimensional data, in order to improve coding efficiency.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Deetlefs in further view of Lefebvre and Yin et al., WO 2023196217 A1 (hereinafter referred to as “Yin”).
Regarding 3, Deetlefs and Lefebvre do not disclose the method of Claim 2, wherein the computational complexity is measured using thousand multiply-accumulate operations ("KMACs") per pixel.
However, Yin discloses that a computational complexity is measured using thousand multiply-accumulate operations ("KMACs") per pixel. (Yin: “The most often used [complexity] indicators are...the number of multiply- and accumulations (MACs) per pixel in units of a thousand (kMac/pixel) ….and floating point operations per second (FLOPS).” ¶ [00025]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre with measuring computational complexity using a KMACs per pixel, as taught by Yin, as a way to determine encoding or decoding computing capacity, in order to improve coding efficiency, as suggested by Yin, ¶ [00025].
Claims 5-8 are rejected under 35 U.S.C. 103 as being unpatentable over Deetlefs in further view of Lefebvre and Besenbruch et al., US 20230154055 A1 (hereinafter referred to as “Besenbruch”).
Regarding claim 5, Deetlefs and Lefebvre do not disclose, the method of Claim 1, wherein the encoding computation budget is at least forty- five times greater than the decoding computation budget.
However, in the same field of endeavor, Besenbruch discloses the encoding computation budget is at least forty- five times greater than the decoding computation budget (Besenbruch: “This asymmetry is reflected by the asymmetric encoding-decoding process of traditional image and video compression approaches. Algorithms such as WebP, HEIC, HEIF, HEVC, AV1 and others, have 100×-1000× runtime differences between encoding and decoding, with decoding being quick and encoding being slow. For instance, … a 100× complexity increase in encoding would be acceptable without causing any problems.” ¶ [1657]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre such that encoding computation budget is at least forty- five times greater than the decoding computation budget, as taught by Besenbruch, without causing problems. Besenbruch, ¶ [1657].
Regarding claim 6, the combination of Deetlefs, Lefebvre and Besenbruch discloses the method of Claim 1, wherein the encoding computation budget is at least thirty- five times greater than the decoding computation budget (Claim 6 is rejected for the same reasons as claim 5, because the claimed encoding computation budget is less than that of claim 5. Besenbruch: “100×-1000× runtime differences between encoding and decoding, with decoding being quick and encoding being slow. For instance, … a 100× complexity increase in encoding would be acceptable without causing any problems.” ¶ [1657]).
Regarding claim 7, the combination of Deetlefs, Lefebvre and Besenbruch discloses the method of Claim 1, wherein the encoding computation budget is at least twenty-five times greater than the decoding computation budget (Claim 7 is rejected for the same reasons as claim 5 because the claimed encoding computation budget is less than that of claim 5. Besenbruch: “100×-1000× runtime differences between encoding and decoding, with decoding being quick and encoding being slow. For instance, … a 100× complexity increase in encoding would be acceptable without causing any problems.” ¶ [1657]).
Regarding claim 8, the combination of Deetlefs, Lefebvre and Besenbruch discloses the method of Claim 1, wherein the encoding computation budget is at least ten times greater than the decoding computation budget (Claim 8 is rejected for the same reasons as claim 5 because the claimed encoding computation budget is less than that of claim 5.) Besenbruch: “100×-1000× runtime differences between encoding and decoding, with decoding being quick and encoding being slow. For instance, … a 100× complexity increase in encoding would be acceptable without causing any problems.” ¶ [1657]).
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Deetlefs in further view of Lefebvre and Tu et al., US 20210182504 A1 (hereinafter referred to as “Tu”).
Regarding claim 14, neither Deetlefs nor Lefebvre disclose the method of Claim 13, wherein the iterative encoding is Stochastic Gumbel Annealing ("SGA").
However, in the same field of endeavor, Tu discloses iterative encoding is Stochastic Gumbel Annealing ("SGA") (Tu: Iterative encoding. ¶ [0034]. “Gumbel-SoftMax” (i.e., Stochastic Gumbel Annealing). ¶ ¶ [0090] and [0097]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre with Stochastic Gumbel Annealing, as taught by Tu, in order to allow the encoder to be trained end-to-end using standard gradient descent while selecting optimal, hard discrete outputs.
Claims 17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Deetlefs in further view of Lefebvre and Racape et al., US 20240155148 A1 (hereinafter referred to as “Racape”).
Regarding claim 17, the combination of Deetlefs and Lefebvre does not disclose the method of Claim 16, wherein the residual convolution transposed layer is implemented with same k1, s1 and N parameters as the first transposed convolution layer.
However, in the same field of endeavor Racape discloses the residual convolution transposed layer being implemented with same k1, s1 and N parameters as the first transposed convolution layer (Racape: 5x5 Trconv layers (i.e., convolution transposed layers) with stride of 2. Fig. 10).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre with the residual convolution transposed layer being implemented with same k1, s1 and N parameters as the first transposed convolution layer, as taught by Racape, in order to improve coding efficiency.
Regarding claim 18, the combination of combination of Deetlefs, Lefebvre and Racape disclose the method of Claim 17, wherein a nonlinear activation is applied to an output of the first transposed convolution layer.
Deetlefs does not disclose a nonlinear activation is applied to an output of the first transposed convolution layer.
However, Lefebvre discloses a nonlinear activation is applied to an output of the first transposed convolution layer (Lefebvre: Activation function “ReLU”. Figs. 4A and 4B).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre with applying a nonlinear activation to an output of the first transposed convolution layer, as taught by Deetlefs, to perform integer convolutions, in order to improve coding efficiency. Deetlefs, ¶ [0061].
Regarding claim 19, the combination of Deetlefs and Lefebvre does not disclose the method of Claim 18, wherein the decoding computation budget remains low by selecting the ki to be greater than the k2, and selecting the si to be greater than the s2.
However, Racape discloses the decoding computation budget remains low by selecting the ki to be greater than the k2, and selecting the si to be greater than the s2 (Racape: 5x5 and 3x3 Trconv layers with strides of 2 and 1. Fig. 10).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Deetlefs and Lefebvre with the decoding computation budget remains low by selecting the ki to be greater than the k2, and selecting the si to be greater than the s2, as taught by Racape, in order to improve coding efficiency.
Allowable Subject Matter
Claims 20, 21 and 24-28 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DEIRDRE L BEASLEY whose telephone number is (571)270-0452. The examiner can normally be reached Monday-Friday 8 a.m. -5 p.m.
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, Chris Kelley can be reached at (571) 272-7331. 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.
/DLB/Patent Examiner, Art Unit 2482
/BEHROOZ M SENFI/Primary Examiner, Art Unit 2482
Prosecution Insights