Qualcomm Incorporated

Cheng-Teh Hsieh

Marta Karczewicz

Vadim Seregin

Wei-Jung Chien

In Suk Chong

Xianglin Wang

Krishnakanth Rapaka

Luong Pham Van

Rajan Laxman Joshi

Joel Sole Rojals

Adarsh Krishnan Ramasubramonian

Geert Van Der Auwera

Chao Pang

Han Huang

Sumit Mohan

Xiang Yang Li

Jianle Chen

Amir Said

Nan Hu

Chun-Chi Chen

Feng Ge

Clara Ka Wah Sung

Xiang Li

Milivoje M Aleksic

Wei Pu

Sungwon Lee

Yung-Hsuan Chao

Serag Gadelrab

Meghal Varia

Jason Edward Podaima

Tom Longo

Ling Feng Huang

Thibaud Laurent Biatek

In-Suk Chong

Natan Jacobson

Yan Zhang

Li Zhang

Victor Szeto

Richard Boisjoly

Li Feng Zhang

Kai Zhang

Ye-Kui Wang

Xin Zhao

Dmytro Rusanovskyy

Vijayaraghavan Thirumalai

Jie Dong

Hilmi Enes Egilmez

Tao Wang

Hsiao-Chiang Chuang

Hongtao Wang

Feng Zou

Yao-Jen Chang

Andrew Edmund Turner

Yu Han

Zhi Zhang

Kevin Pascal Andre Reuze

Jaehong Chon

Ankitesh Kumar Singh

Stephen Cheung

Eric Michael Lutz

Ravikanth Challagulla

Rohan Prakash Mallya

Hyung Joon Kim

Systems, devices, apparatus, and methods, including computer programs encoded on storage media, for truncation error signaling and adaptive dither for lossy bandwidth compression are described. A processor may perform a truncation process for data, where the data is associated with display processing, image processing, or the data processing, where the truncation process for the data results in truncated data. The processor may compute a set of truncation error values associated with the truncation process for the truncated data. The processor may generate a set of residual samples for the truncated data. The processor may generate a bitstream based on the set of residual samples for the truncated data and the set of truncation error values associated with the truncation process.

Truncation error signaling and adaptive dither for lossy bandwidth compression

A video decoder may be configured to receive a block of video data that was encoded using a coding mode that includes a search process in one or more reference frames. The video decoder may prefetch reference samples in a fixed search region of at least one reference frame of the one or more reference frames, and decode the block of video data using the coding mode, including performing the search process for the coding mode using the prefetched reference samples.

Searching techniques for template matching for video coding

A device for decoding video data includes a memory configured to store video data; and one or more processors implemented in circuitry and configured to: determine a deterministic bounding box from which to retrieve reference samples of reference pictures of video data for performing decoder-side motion vector derivation (DMVD) for a current block of the video data; derive a motion vector for the current block according to DMVD using the reference samples within the deterministic bounding box; form a prediction block using the motion vector; and decode the current block using the prediction block.

Using unrefined motion vectors for performing decoder-side motion vector derivation

An example method includes determining, for each respective coding block of a plurality of coding blocks of a current coding tree unit (CTU) of video data in a current picture of video data, a respective search area of a plurality of respective search areas, wherein at least one of the plurality of search areas includes samples of the current picture located outside of the current CTU, and wherein at least one of the plurality of search areas does not include samples of the current picture located outside of the current CTU; selecting, for each respective coding block and from within the respective search area for the respective coding block, a respective predictor block of a plurality of predictor blocks; and reconstructing samples of each respective coding block based on samples included in a corresponding predictor block in the plurality of predictor blocks.

Intra block copy prediction restrictions in video coding

The present disclosure relates to methods and devices for data or graphics processing including an apparatus, e.g., a GPU. The apparatus may receive at least one bitstream including a plurality of bits, each of the bits corresponding to a position in the at least one bitstream, and each of the bits being associated with color data. The apparatus may also arrange an order of the plurality of bits in the at least one bitstream, such that at least one of the bits corresponds to an updated position in the at least one bitstream. Additionally, the apparatus may convert, upon arranging the order of the bits, the color data associated with each of the plurality of bits in the at least one bitstream. The apparatus may also compress, upon converting the color data associated with each of the bits, the plurality of bits in the at least one bitstream.

Methods and apparatus for lossless compression of GPU data

An example device for encoding high dynamic range (HDR) video data includes a memory configured to store video data; and one or more processors implemented in circuitry and configured to: calculate a histogram for an image of the video data, the image being expressed in a linear light format; encode values for the histogram of the image expressed in the linear light format; and encode the image. Data for the histogram may be expressed in an array of variables having a size of 2×18 bits. The device may encode codewords representing values for bins of the histogram, where the codewords may be selected from a set of codewords for a PQ10 format for HDR images. The bins of the histogram may represent non-equal width ranges.

Encoding high dynamic range video data

An example device for decoding video data includes one or more processors implemented in circuitry and configured to: decode a coding tree unit (CTU) of video data, the CTU including a luminance (luma) block and a chrominance (chroma) block, to produce a decoded luma block and a decoded chroma block; determine that a chroma sample of the decoded chroma block is on a first side of an adaptive loop filter (ALF) virtual boundary and that a co-located luma sample of the decoded luma block is on a second side of the ALF virtual boundary, the co-located luma sample being co-located with the chroma sample, the first side being different than the second side; and in response to determining that the chroma sample is on the first side and the luma sample is on the second side, disable cross-component adaptive loop filtering (CC-ALF) for the chroma sample.

Cross-component adaptive loop filter in video coding

A video coder may be configured to code video data by performing splitting of a coding unit (CU) of video data using intra sub-partition (ISP) to form a set of prediction blocks. The video coder may group a plurality of the prediction blocks from the set of prediction blocks into a first prediction block group (PBG). The video coder may reconstruct samples of prediction blocks included in the first PBG independently of samples of other prediction blocks included in the first PBG.

Subblock coding by generalized intra prediction in video coding

A video coder determines a plurality of available Matrix Intra Prediction (MIP) parameter sets (MPS's) for a picture of video data. The plurality of available MPS's is a union of (i) a subset of all default MPS's and (ii) a set of additional MPS's that are signaled in the bitstream. Each of the default MPS's is associated with a predefined MIP mode in a codec. Each of the set of additional MPS's is associated with a new MIP mode in a set of new MIP modes. The video decoder uses a MIP mode associated with an MPS in the plurality of available MPS's to generate a prediction block for a current block of the picture.

Signaling of Matrix Intra Prediction parameters in video coding

Techniques and systems are provided for compressing data in a neural network. For example, output data can be obtained from a node of the neural network. Re-arranged output data having a re-arranged scanning pattern can be generated. The re-arranged output data can be generated by re-arranging the output data into the re-arranged scanning pattern. One or more residual values can be determined for the re-arranged output data by applying a prediction mode to the re-arranged output data. The one or more residual values can then be compressed using a coding mode.

Bandwidth compression for neural network systems

Techniques for processing video data are described. The techniques include determining a first partition and a second partition for a current block coded in geometric partition mode, determining a first and second prediction block based on a first and second motion vector, blending the first prediction block and the second prediction block based on weights indicative of an amount to scale samples in the first prediction block and the second prediction block to generate a final prediction block, dividing the current block into a plurality of sub-blocks, determining a set of sub-blocks that each include at least one sample that corresponds to a prediction sample in the final prediction block that was generated based on equal weighting of a sample in the first prediction block and a sample in the second prediction block, and storing respective bi-prediction motion vectors for each sub-block in the determined set of sub-blocks.

Geometric partition mode with harmonized motion field storage and motion compensation

Techniques are described of hybrid coders that are configured to selectively use adaptive or non-adaptive coding techniques. A video coder (e.g., video encoder or video decoder) may code (e.g., encode or decode) first video data (e.g., a syntax element or value), for coding a first block, based on an adaptive context model (e.g., one or more adaptive context models) and code second video data, for coding a second block, based on a non-adaptive context model (e.g., one or more non-adaptive context models).

Arithmetic coding with selective adaptation for video coding

Various embodiments include methods and devices for implementing decompression of compressed high dynamic ratio fields. Various embodiments may include receiving compressed first and second sets of data fields, decompressing the first and second compressed sets of data fields to generate first and second decompressed sets of data fields, receiving a mapping for mapping the first and second decompressed sets of data fields to a set of data units, aggregating the first and second decompressed sets of data fields using the mapping to generate a compression block comprising the set of data units.

Compression of high dynamic ratio fields for machine learning

Growing community of inventors

Del Mar, CA, United States of America

Cheng-Teh Hsieh