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 .
Response to Arguments
Applicant’s arguments, see Response to Office Action mailed 3 February 2026, filed 1 May 2026, with respect to Claim Objections and Claim Rejections under 35 USC §112 have been fully considered and are persuasive. The Claim Objections and Claim Rejections under 35 USC §112 of claims 3 and 18 has been withdrawn.
Applicant's arguments filed 1 May 2026 regarding Claim Rejections under 35 USC §102 and §103 have been fully considered but they are not persuasive.
Applicant argues the primary reference, Cho, does not disclose all limitations of the claims. Applicant argues “According to claim 3, two different entities of data are encoded into the data stream:
- The set of features, which is obtained using the first machine learning predictor, and which represents a motion estimation for the picture to be encoded; and
- The residual picture derived based on a motion-predicted picture and the picture to be encoded.”
Examiner respectfully disagrees and respectfully directs Applicant’s attention to Cho: ¶ [0683] The input of the generation encoder may be a residual frame. The generation encoder may generate the feature vector of the input residual frame.
and
¶ [0684] The generation decoder may generate a future frame by adding the residual frame to a previously reconstructed frame. The generation decoder may generate an added feature vector by adding the feature vector of the previously reconstructed frame to the predicted feature vector of the residual frame. The generation decoder may generate the future frame using the added feature vector.
Examiner respectfully submits both the residual frame and feature vector disclose the two different entities of data which are encoded into the data stream as show in Cho: FIG. 20.
Claim Rejections - 35 USC § 102
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 3-7, 18, and 20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Cho et al. (US 2019/0246102 A1).
Regarding Claims 3 and 18, Cho discloses an apparatus and method for encoding a picture of a video into encoding a picture of a video into a data stream [Cho: FIG. 1], configured for using a first machine learning predictor to derive a set of features representing a motion estimation for the picture with respect to a previous picture of the video [Cho: ¶ [0010]: The feature vector of the target frame may be generated by a first convolutional neural network.
¶ [0011]: The residual frame may be acquired through motion prediction that uses a motion vector of the target frame; and ¶ [0012]: The feature vector of the residual frame may be generated by a second convolutional neural network], encoding the set of features into the data stream [Cho: ¶ [0010]-[0012]; ¶ [0683] The input of the generation encoder may be a residual frame. The generation encoder may generate the feature vector of the input residual frame; ¶ [0684] The generation decoder may generate a future frame by adding the residual frame to a previously reconstructed frame. The generation decoder may generate an added feature vector by adding the feature vector of the previously reconstructed frame to the predicted feature vector of the residual frame. The generation decoder may generate the future frame using the added feature vector; and FIG. 20], predicting the picture using the set of features to derive a residual picture, by using a second machine learning predictor to determine a set of reconstructed motion vectors based on the features [Cho: ¶ [0013]: The predicted feature vector for the residual frame may be generated by a convolution Long Short Term Memory (LSTM) neural network] , deriving a motion-predicted picture based on the previous picture using the set of reconstructed motion vectors [Cho: ¶ [0014]: The prediction frame may be generated by a deconvolutional neural network, and ¶ [0015]: In accordance with another aspect, there is provided a prediction method, including generating a virtual frame; and performing inter prediction that uses the virtual frame, wherein the virtual frame is generated using a neural network to which a previously decoded frame is input], and deriving the residual picture based on the motion-predicted picture and the picture [Cho: ¶ [0016]: The virtual frame may be generated based on the previously decoded frame, a residual frame, and a reconstructed prediction residual frame], and encoding the residual picture into the data stream [Cho: ¶ [0684] The generation decoder may generate a future frame by adding the residual frame to a previously reconstructed frame. The generation decoder may generate an added feature vector by adding the feature vector of the previously reconstructed frame to the predicted feature vector of the residual frame. The generation decoder may generate the future frame using the added feature vector], wherein the apparatus is configured for optimizing the features with respect to a rate-distortion measure for the features, the rate-distortion measure being determined based on a distortion between the picture and the motion- predicted picture [Cho: ¶ [0140]: Rate-distortion optimization: An encoding apparatus may use rate-distortion optimization so as to provide high coding efficiency by utilizing combinations of the size of a coding unit (CU), a prediction mode, the size of a prediction unit (PU), motion information, and the size of a transform unit (TU)].
Regarding Claim 4, Cho discloses all the limitations of Claim 3, and is analyzed as previously discussed with respect to that claim.
Furthermore, Cho discloses configured for quantizing the features to acquire quantized features [Cho: ¶ [0174]: Quantization Parameter (QP): A quantization parameter may be a value used to generate a transform coefficient level for a transform coefficient in quantization. Alternatively, a quantization parameter may also be a value used to generate a transform coefficient by scaling the transform coefficient level in dequantization. Alternatively, a quantization parameter may be a value mapped to a quantization step size], and determining the set of reconstructed motion vectors using the second machine learning predictor based on the quantized features [Cho: ¶ [0178]: A quantized level or a quantized transform coefficient level generated by applying quantization to a transform coefficient or a residual signal may also be included in the meaning of the term “transform coefficient”].
Regarding Claim 5, Cho discloses all the limitations of Claim 3, and is analyzed as previously discussed with respect to that claim.
Furthermore, Cho discloses configured for optimizing the features using a gradient descent algorithm with respect to the rate-distortion measure [Cho: ¶ ¶ [0739]: In the learning process by the RNN, a vanishing gradient problem, in which previously input data (i.e. past data) vanishes with the lapse of time, may occur. The LSTM may be used to solve the vanishing gradient problem. The structure of the LSTM allows the gradient of errors to propagate backwards in time in the neural network. In other words, the structure of the LSTM may be configured such that data previously input to the neural network influences the current output of the neural network, either more continuously or more strongly].
Regarding Claim 7, Cho discloses all the limitations of Claim 3, and is analyzed as previously discussed with respect to that claim.
Furthermore, Cho discloses wherein the second machine learning predictor comprises a convolutional neural network comprising a plurality of linear convolutional layers using rectifying linear units as activation functions [Cho: ¶ [0058]: FIG. 25 illustrates an operation in a Rectified Linear Unit (ReLu) layer according to an example; and ¶ [0697]: The first CNN of the generation encoder may include a convolution layer, a pooling layer, and a Rectified Linear Unit (ReLu) layer. The convolution layer, the pooling layer, and the ReLu layer may each include multiple layers] and/or wherein the second machine learning predictor comprises a linear transfer function.
Regarding Claim 20, Cho discloses all the limitations of Claim 3, and is analyzed as previously discussed with respect to that claim.
Furthermore, Cho discloses configured for encoding the set of features into the data stream by quantizing the set of features to obtain quantized features, and encoding the quantized features into the data stream [Cho: ¶ [0211] The residual signal may be the difference between an original signal and a prediction signal. Alternatively, the residual signal may be a signal generated by transforming or quantizing the difference between an original signal and a prediction signal or by transforming and quantizing the difference. A residual block may be a residual signal for a block unit].
Claim Rejections - 35 USC § 103
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Cho as applied to claim 3 above, and further in view of Bendapudi et al. (US 2023/0108645 A1).
Regarding Claim 6, Cho discloses all the limitations of Claim 3, and is analyzed as previously discussed with respect to that claim.
Cho may not explicitly disclose configured for determining a rate measure for the rate-distortion measure based on the residual picture using a spatial-to-spectral transformation, and/or determining the distortion between the picture and the motion-predicted picture based on the residual picture using a spatial-to-spectral transformation.
However, Bendapudi discloses configured for determining a rate measure for the rate-distortion measure based on the residual picture using a spatial-to-spectral transformation, and/or determining the distortion between the picture and the motion-predicted picture based on the residual picture using a spatial-to-spectral transformation [Bendapudi: ¶ [0095] As part of residual coding, in the transformer/scaler/quantizer (530), when a frequency transform is not skipped, a frequency transformer converts spatial-domain video information into frequency-domain (i.e., spectral, transform) data].
It would have been obvious to one having ordinary skill in the art before the effective filing date to combine the transformation of Bendapudi with the processing of Cho in order to provide processing on more types of data, improving usability.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JONATHAN R MESSMORE whose telephone number is (571)272-2773. The examiner can normally be reached Monday-Friday 9-5 EST/EDT.
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.
/JONATHAN R MESSMORE/Primary Examiner, Art Unit 2482
Prosecution Insights