Patents by Inventor Madhu Peringassery Krishnan
Madhu Peringassery Krishnan has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11930186Abstract: Aspects of the disclosure provide a method and an apparatus including processing circuitry for video decoding. The processing circuitry decodes coded information for a transform block (TB) from a coded video bitstream. The coded information indicates one of intra prediction mode information that indicates an intra prediction mode used for the TB, a size of the TB, and a primary transform type used for the TB. The processing circuitry determines a context for entropy decoding a secondary transform index based on the one of the intra prediction mode information for the TB, the size of the TB, and the primary transform type used for the TB. The secondary transform index indicates a secondary transform in a set of secondary transforms that is to be performed on the TB. The processing circuitry entropy decodes the secondary transform index based on the context and performs the secondary transform.Type: GrantFiled: June 28, 2021Date of Patent: March 12, 2024Assignee: TENCENT AMERICA LLCInventors: Madhu Peringassery Krishnan, Xin Zhao, Shan Liu
-
Patent number: 11930177Abstract: This disclosure relates generally to video coding and particularly to cross component level reconstruction. For example, a method is disclosed for processing video data which may include extracting a first transform block of a first color component and a second transform block of a second color component from a bitstream of a video block; determining that transform coefficients in the first transform block are all zero; determining that a CCLR is applied to the first transform block; refining one or more of the transform coefficients in the first transform block, to obtain a refined first transform block; determining a target transform kernel for the refined first transform block; performing a reverse transform on the refined first transform block based on the target transform kernel to obtain a target block; and reconstructing the first color component of the video block based on at least the target block.Type: GrantFiled: October 14, 2022Date of Patent: March 12, 2024Assignee: Tencent America LLCInventors: Madhu Peringassery Krishnan, Xin Zhao, Shan Liu
-
Publication number: 20240080446Abstract: The various embodiments described herein include methods and systems for encoding and decoding video. In one aspect, a method includes receiving a video bitstream that includes a plurality of blocks. The method also includes obtaining, from the video bitstream, respective values for one or more parameters associated with a block in the plurality of blocks, the one or more parameters corresponding to an arithmetic coder probability update rate, where the respective values for the one or more parameters are determined from a set of predefined values. The method further includes calculating the arithmetic coder probability update rate based on the respective values for the one or more parameters signaled in the video bitstream, and determining one or more coding contexts based on the calculated arithmetic coder probability update rate. The method also includes decoding the block in the plurality of blocks based on the one or more coding contexts.Type: ApplicationFiled: June 13, 2023Publication date: March 7, 2024Inventors: Madhu Peringassery KRISHNAN, Xin ZHAO, Shan LIU
-
Patent number: 11917202Abstract: A method and apparatus for performing a frequency-dependent joint component secondary transform (FD-JCST). The method includes obtaining a plurality of transform coefficients in a transform coefficient block; determining whether at least one of the plurality of transform coefficients is a low-frequency coefficient; based on determining that the at least one of the plurality of transform coefficients is the low-frequency coefficient, determining whether the low-frequency coefficient is a non-zero value; and based on determining that the low-frequency coefficient is the non-zero value, performing a joint component secondary transform (JCST) on the low-frequency coefficient and signaling a related syntax to indicate that the JCST is performed.Type: GrantFiled: October 6, 2021Date of Patent: February 27, 2024Assignee: TENCENT AMERICA LLCInventors: Xin Zhao, Madhu Peringassery Krishnan, Shan Liu
-
Patent number: 11882301Abstract: A method, computer program, and computer system is encoding or decoding video data. Video data may include a syntax element indicating a quantization index, wherein a range of the quantization index is extended by an offset value. The syntax element may be parsed to determine the quantization index, a quantization step size may be obtained from a lookup table based on the quantization index and the offset value; and the video data may be encoded or decoded based on the quantization step size.Type: GrantFiled: June 10, 2022Date of Patent: January 23, 2024Assignee: TENCENT AMERICA LLCInventors: Xin Zhao, Madhu Peringassery Krishnan, Shan Liu
-
Publication number: 20240015303Abstract: An approach for encoding/decoding video data, performed by at least one processor, including: obtaining video data; parsing the video data into blocks, wherein the blocks are associated with reference picture lists; generating a warp model used for a first reference picture list and a second reference picture list of a current block, included in the reference picture lists, based on motion vectors of the current block and adjacent blocks to the current block; decoding a frame among the first reference picture list and the second reference picture list by applying the warp model to the frame.Type: ApplicationFiled: November 9, 2022Publication date: January 11, 2024Applicant: TENCENT AMERICA LLCInventors: Han GAO, Xin ZHAO, Liang ZHAO, Madhu PERINGASSERY KRISHNAN, Shan LIU
-
Publication number: 20240015331Abstract: Systems and methods for encoding and decoding using syntax design for multi-symbol arithmetic coding are provided. A method includes receiving a coded video bitstream including a plurality of syntax elements; determining a first maximum alphabet size for arithmetic coding by an arithmetic coding engine, the first maximum alphabet size determined based on a hardware constraint; determining a second maximum alphabet size that is less than the first maximum alphabet size; and decoding the plurality of syntax elements included in the coded video bitstream, based on the determined second maximum alphabet size, wherein each of the plurality of syntax elements is entropy coded with an alphabet size less than or equal to the determined second maximum alphabet size.Type: ApplicationFiled: November 9, 2022Publication date: January 11, 2024Applicant: TENCENT AMERICA LLCInventors: Madhu PERINGASSERY KRISHNAN, Liang ZHAO, Xin ZHAO, Jing YE, Shan LIU
-
Publication number: 20240015278Abstract: A method includes receiving a current picture, a first reference picture, and a second reference picture. The method includes obtaining a plurality of predefined weighting patterns, each weighting pattern being signaled as an index value and selecting a weighting pattern based on a predetermined condition. The method includes deriving a first weight to be applied to a first sub-block in the first reference picture and a second weight to be applied to a second sub-block in the second reference picture based on the index value corresponding to the selected weighting pattern. The method includes assigning the first weight to the first sub-block and the second weight to the second sub-block based on the selected weighting pattern. The method includes decoding the current block by a weighted bi-prediction based at least on the first sub-block weighted by the first weight and the second sub-block weighted by the second weight.Type: ApplicationFiled: November 3, 2022Publication date: January 11, 2024Applicant: Tencent America LLCInventors: Madhu PERINGASSERY KRISHNAN, Xin Zhao, Liang Zhao, Han Gao, Xiaozhong Xu, Shan Liu
-
Publication number: 20240007673Abstract: This disclosure relates to a transform kernel sharing in video encoding and decoding. For example, a method is disclosed for such transform kernel sharing. The method may include identifying a plurality of transform kernels, wherein each of the plurality of transform kernels comprises a set of basis vectors from low to high frequencies; N high-frequency basis vectors of two or more of the plurality of transform kernels are shared, N being a positive integer; and low-frequency basis vectors of the two or more of the plurality of the transform kernels other than the N high-frequency basis vectors are individualized. The method may further include extracting a data block from a video bitstream; selecting a transform kernel from the plurality of transform kernels based on information associated with the data block; and applying the transform kernel to at least a portion of the data block to generate a transformed block.Type: ApplicationFiled: September 12, 2023Publication date: January 4, 2024Applicant: TENCENT AMERICA LLCInventors: Xin ZHAO, Madhu PERINGASSERY KRISHNAN, Shan LIU
-
Patent number: 11863775Abstract: This disclosure relates to cross component methods for refining decoded transform coefficients before or after dequantization in video decoding. For example, a method for video decoding is disclosed. The method may include, comprising extracting a first transform coefficient of a first color component from a bitstream of a coded video; extracting a second transform coefficient of a second color component from the bitstream of the coded video; deriving an offset value based on a magnitude or sign value of the first transform coefficient; adding the offset value to a magnitude of the second transform coefficient to generate a modified second transform coefficient for the second color component; and reconstructing the coded video based on at least the first transform coefficient of the first color component and the modified second transform coefficient of the second color component.Type: GrantFiled: January 13, 2022Date of Patent: January 2, 2024Assignee: Tencent America LLCInventors: Xin Zhao, Madhu Peringassery Krishnan, Shan Liu
-
Patent number: 11863790Abstract: A method of decoding may be performed by at least one processor, and may comprise: receiving an entropy coded bitstream comprising compressed video data; generating one or more dequantized blocks, determining whether at least one of a height and a width of the one or more dequantized blocks is greater than or equal to a predefined threshold, and responsive to the at least one of the height or the width of the one or more dequantized blocks being greater than or equal to the predefined threshold, transform coding a dequantized block using a tuned line graph transform (LGT) core to perform direct matrix multiplications for each of the horizontal and vertical dimensions of the one or more dequantized blocks.Type: GrantFiled: January 10, 2022Date of Patent: January 2, 2024Assignee: TENCENT AMERICA LLCInventors: Madhu Peringassery Krishnan, Xin Zhao, Shan Liu
-
Publication number: 20230421754Abstract: A method, computer program, and computer system is provided for coding video data. Reference samples and magnitudes of transform coefficients corresponding to a current block of video data from an input to a neural network are identified. Sign values associated with the transform coefficients are predicted using neural networks. The video data is encoded/decoded based on the predicted sign values.Type: ApplicationFiled: September 11, 2023Publication date: December 28, 2023Applicant: TENCENT AMERICA LLCInventors: Xin ZHAO, Yixin DU, Liang ZHAO, Madhu PERINGASSERY KRISHNAN, Shan LIU
-
Patent number: 11856225Abstract: A method of decoding image data is provided. The method may include generating a dequantized block of an image based on a coded bitstream; determining whether to use one from among an implicit method and an explicit method for selecting a hybrid transform kernel from among one or more hybrid transform kernel that are available for decoding the dequantized block, wherein the one or more hybrid transform kernel are available for decoding the dequantized block based on associations between prediction modes and sizes of blocks; selecting, by using the one from among the implicit method and the explicit method, the hybrid transform kernel from among the one or more hybrid transform kernel; and performing inverse transform coding of the dequantized block based on the selected hybrid transform kernel.Type: GrantFiled: October 28, 2022Date of Patent: December 26, 2023Assignee: TENCENT AMERICA LLCInventors: Madhu Peringassery Krishnan, Xin Zhao, Shan Liu
-
Publication number: 20230412812Abstract: The various embodiments described herein include methods and systems for encoding and decoding video. In one aspect, a method includes obtaining video data that includes a first block. The method further includes determining a plurality of transform coefficients associated with the first block. The method also includes predicting respective coefficient signs for the plurality of transform coefficients. The method also includes entropy coding the first block including jointly signaling two or more of the predicted coefficient signs.Type: ApplicationFiled: May 4, 2023Publication date: December 21, 2023Inventors: Samruddhi Yashwant KAHU, Madhu PERINGASSERY KRISHNAN, Xin ZHAO, Shan LIU
-
Patent number: 11849115Abstract: A method includes determining whether at least one of a height or width of a residual coding block, that corresponds to a data block of an image, is greater than or equal to a threshold; and based on determining that the at least one of the height or width of the residual coding block is greater than or equal to the threshold: obtaining a reduced-complexity residual coding block based on the residual coding block and performing transform coding of the reduced-complexity residual coding block using a line graph transform (LGT) core to perform direct matrix multiplications for each of the horizontal and vertical dimensions of the reduced-complexity residual coding block, wherein the performing includes determining an order of performing horizontal and vertical transforms to the reduced-complexity residual coding block when zero-out is enabled on LGT, depending on a ratio of width to height of the residual coding block.Type: GrantFiled: March 10, 2022Date of Patent: December 19, 2023Assignee: TENCENT AMERICA LLCInventors: Madhu Peringassery Krishnan, Xin Zhao, Shan Liu
-
Publication number: 20230370603Abstract: The various embodiments described herein include methods and systems for coding video. In one aspect, a method includes obtaining video data comprising a plurality of blocks, including a first block and determining a plurality of transform coefficients associated with the first block. The method further includes predicting, using a first technique, respective coefficient signs for a first set of the plurality of transform coefficients and predicting, using a second technique, respective coefficient signs for a second set of the plurality of transform coefficients, the second technique being different than the first technique. The method also includes reconstructing the first block based on the plurality of transform coefficients and the predicted respective coefficient signs for the first and second sets.Type: ApplicationFiled: March 28, 2023Publication date: November 16, 2023Inventors: Xin ZHAO, Samruddhi Yashwant KAHU, Madhu PERINGASSERY KRISHNAN, Shan LIU
-
Publication number: 20230336727Abstract: A system includes code configured to cause a processor to obtain a video bitstream, the video bitstream including: a first quantization index value for a coefficient of a coded image; an offset value; a quantization step size that corresponds to the first quantization index value; a second quantization index value for another coefficient of the coded image, the second quantization index value being based on both the first quantization index value and the offset value and being greater than or equal to a predetermined threshold value; and a mode indicating whether the coded image is to be decoded in a lossy mode or a lossless mode, the mode being determined based on whether the first quantization index value is equal to a quantization index value associated with lossless coding, and whether the offset value is less than or equal to the quantization index value associated with the lossless coding.Type: ApplicationFiled: June 27, 2023Publication date: October 19, 2023Applicant: TENCENT AMERICA LLCInventors: Madhu Peringassery KRISHNAN, Xin Zhao, Shan Liu
-
Patent number: 11792431Abstract: This disclosure relates to a transform kernel sharing in video encoding and decoding. For example, a method is disclosed for such transform kernel sharing. The method may include identifying a plurality of transform kernels, wherein each of the plurality of transform kernels comprises a set of basis vectors from low to high frequencies; N high-frequency basis vectors of two or more of the plurality of transform kernels are shared, N being a positive integer; and low-frequency basis vectors of the two or more of the plurality of the transform kernels other than the N high-frequency basis vectors are individualized. The method may further include extracting a data block from a video bitstream; selecting a transform kernel from the plurality of transform kernels based on information associated with the data block; and applying the transform kernel to at least a portion of the data block to generate a transformed block.Type: GrantFiled: January 5, 2022Date of Patent: October 17, 2023Assignee: Tencent America LLCInventors: Xin Zhao, Madhu Peringassery Krishnan, Shan Liu
-
Patent number: 11792391Abstract: A method, computer program, and computer system is provided for coding video data. Reference samples and magnitudes of transform coefficients corresponding to a current block of video data from an input to a neural network are identified. Sign values associated with the transform coefficients are predicted based on at least the identified reference samples. The video data is encoded/decoded based on the predicted sign values.Type: GrantFiled: October 5, 2021Date of Patent: October 17, 2023Assignee: TENCENT AMERICA LLCInventors: Xin Zhao, Yixin Du, Liang Zhao, Madhu Peringassery Krishnan, Shan Liu
-
Publication number: 20230319260Abstract: The various embodiments described herein include methods and systems for coding video. In one aspect, a method includes obtaining video data comprising a plurality of blocks, including a first block. The method further includes determining a plurality of transform coefficients associated with the first block; and selecting a subset of the plurality of transform coefficients based on a frequency corresponding to each transform coefficient of the plurality of transform coefficients. The method also includes predicting respective coefficient signs for selected subset of the transform coefficients associated with the first block based on respective context information; and reconstructing the first block based on the plurality of transform coefficients and the predicted respective coefficient signs.Type: ApplicationFiled: March 14, 2023Publication date: October 5, 2023Inventors: Samruddhi Yashwant KAHU, Xin ZHAO, Madhu PERINGASSERY KRISHNAN, Shan LIU