Qualcomm Incorporated

Velos Media International Limited

Muhammed Zeyd Coban

Marta Karczewicz

Vadim Seregin

Geert Van Der Auwera

Adarsh Krishnan Ramasubramonian

Yong He

Peisong Chen

Hongtao Wang

Hilmi Enes Egilmez

Yao-Jen Chang

Wei-Jung Chien

Rajan Laxman Joshi

Alican Nalci

Yunfei Zheng

Ye-Kui Wang

Xianglin Wang

Yung-Hsuan Chao

Joel Sole Rojals

Nan Hu

Fnu Hendry

Jie Dong

Jianle Chen

Ying Hai Chen

Hsiao-Chiang Chuang

Yu Han

Nikolai Konrad Leung

Ankitesh Kumar Singh

Alberto Rico Alvarino

Umesh Phuyal

Thomas Stockhammer

Bappaditya Ray

Luong Pham Van

Louis Joseph Kerofsky

Sungwon Lee

Yang Yu

Siping Tao

Liangping Ma

Yi-Wen Chen

Min Wang

Leena Zacharias

Han Huang

Dmytro Rusanovskyy

Imed Bouazizi

Liwei Guo

In Suk Chong

Seyfullah Halit Oguz

Suhail Jalil

Yan Zhang

Chengjie Tu

Amit Rohatgi

Rahul P Panchal

Junchen Du

Gurunath Ramaswamy

Bo Zhou

Zhen Liu

Hendry

Karthic Veera

Samuel James Eadie

Ying Chen

An example device for filtering video data includes a memory configured to store video data; and a processing system comprising one or more processors implemented in circuitry, the processing system being configured to: apply one or more neural network processing blocks to intermediate filtered video data, each of the neural network processing blocks including a first 1×1 convolutional filter, a parametric rectified linear unit (PReLU) filter, a second 1×1 convolutional filter, and a 3×3 convolutional filter; apply additional neural network processing blocks to output of the one or more neural network processing blocks to form filtered video data; and output the filtered video data.

Reduced complexity multi-mode neural network filtering of video data

A method for coding a block of video data using affine mode includes determining a refined affine model for the current block of video data from a linear regression process using a base motion vector field and a guidance motion vector field as inputs to the linear regression process. The method further includes determining affine merge candidates for the current block using the refined affine model, and coding the current block of video data using the affine merge candidates.

Derivation of affine merge candidates with linear regression for video coding

A video coder is configured to determine a reference block of a reference picture for prediction of a current block of a current picture using motion information and to generate a set of reference samples for the current block of the current picture. To generate the set of reference samples, the video coder is configured to perform reference sample clipping on the reference block of the reference picture based on a size of the reference picture. The video coder is further configured to generate a prediction block for the current block of the current picture based on the set of reference samples.

Motion compensation using size of reference picture

Systems and techniques are described herein for processing video data. For instance, a process can include obtaining video data including a picture. The process can also include determining a width of a film grain synthesis block of the picture based on at least one of a width and a height of the picture. The process can further include determining a height of the film grain synthesis block of the picture is one. The process may also include determining a block size of the film grain synthesis block based on the determined width and height. The process may further include selecting a grain block based on the determined block size.

Adaptive film grain synthesis

A video encoder may transform residual data by using a transform selected from a group of transforms. The transform is applied to the residual data to create a two-dimensional array of transform coefficients. A scanning mode is selected to scan the transform coefficients in the two-dimensional array into a one-dimensional array of transform coefficients. The combination of transform and scanning mode may be selected from a subset of combinations that is based on an intra-prediction mode. The scanning mode may also be selected based on the transform used to create the two-dimensional array. The transforms and/or scanning modes used may be signaled to a video decoder.

Video coding using mapped transforms and scanning modes

A video decoder determines, based on a block size of a current block and a low-frequency non-separable transform (LFNST) syntax element, a zero-out pattern of normatively defined zero-coefficients. The LFNST syntax element is signaled at a transform unit (TU) level. Additionally, the video decoder determines transform coefficients of the current block. The transform coefficients of the current block include transform coefficients in an LFNST region of the current block and transform coefficients outside the LFNST region of the current block. As part of determining the transform coefficients of the current block, the video decoder applies an inverse LFNST to determine values of one or more transform coefficients in the LFNST region of the current block. The video decoder also determines that transform coefficients of the current block in a region of the current block defined by the zero-out pattern are equal to 0.

Transform and last significant coefficient position signaling for low-frequency non-separable transform in video coding

Techniques are described herein for processing video data using a neural network system. For instance, a process can include generating, by a first convolutional layer of an encoder sub-network of the neural network system, output values associated with a luminance channel of a frame. The process can include generating, by a second convolutional layer of the encoder sub-network, output values associated with at least one chrominance channel of the frame. The process can include generating, by a third convolutional layer based on the output values associated with the luminance channel of the frame and the output values associated with the at least one chrominance channel of the frame, a combined representation of the frame. The process can further include generating encoded video data based on the combined representation of the frame.

Front-end architecture for neural network based video coding

A video coder may determine a quantization parameter (QP) value for a block of video data, determine a residual coding method from a plurality of residual coding methods based on the QP value, wherein the plurality of residual coding methods include transform skip (TS) residual coding and regular residual coding, and code a residual of the block of video data using the determined residual coding method.

Residual coding selection and low-level signaling based on quantization parameter

An example device includes memory and one or more processors implemented in circuitry and communicatively coupled to the memory. The one or more processors are configured to receive a first slice header syntax element for a slice of the video data and determine a first value for the first slice header syntax element, the first value being indicative of whether dependent quantization is enabled. The one or more processors are configured to receive a second slice header syntax element for the slice of the video data and determine a second value for the second slice header syntax element, the second value being indicative of whether sign data hiding is enabled. The one or more processors are configured to determine whether transform skip residual coding is disabled for the slice based on the first value and the second value and decode the slice based on the determinations.

High-level constraints for transform skip blocks in video coding

A device for encoding video data includes a memory configured to store video data, and a video encoder implemented in circuitry and configured to encode a future picture of the video data having a first display order position, the future picture being included in an intra period (IP) of the video data, the IP comprising a plurality of groups of pictures (GOPs), and after encoding the future picture, encode a picture of an ordinal first GOP of the plurality of GOPs using the future picture as a reference picture, each picture of the ordinal first GOP having display order positions earlier than the first display order position. Encoding the future picture in this manner may result in encoding performance improvements with minimal increases in encoding and decoding complexity.

Adaptive GOP structure with future reference frame in random access configuration for video coding

A video decoder may apply a first inverse residual modification function to first decoded modified chroma residual data to generate first inverse modified chroma residual data. Additionally, the video decoder may apply a second inverse residual modification function to second decoded modified chroma residual data to generate second inverse modified chroma residual data. The first decoded modified chroma residual data is associated with a first chroma component and the second decoded modified chroma residual data is associated with a second chroma component. The video decoder may reconstruct a block of video data based on the first inverse modified chroma residual data and the second inverse modified chroma residual data.

Combined residual coding in video coding

Various embodiments may provide methods, systems, and devices for supporting application level discovery of Radio Access Network (RAN) statistics and/or events. Various embodiments may provide methods, systems, and devices for supporting application level signaling of Quality of Service (QoS) requirements.

Interface between a radio access network and an application

Systems and techniques are described herein for processing video data. In some aspects, a method can include obtain, by a machine learning system, input video data. The input video data includes one or more luminance components for a current frame. The method can include determining, by the machine learning system, motion information for the luminance component(s) of the current frame and motion information for one or more chrominance components of the current frame using the luminance component(s) for the current frame. In some cases, the method can include determining the motion information for the luminance component(s) based on the luma component(s) of the current frame and at least one reconstructed luma component of a previous frame. In some cases, the method can further include determining the motion information for the chrominance component(s) of the current frame using the motion information determined for the luminance component(s) of the current frame.

Machine learning based flow determination for video coding

An example method of decoding video data that includes receiving one or more syntax elements of the video data indicative of whether a first type of coding scheme or a second type of coding scheme is applied to residual values of a block of video data coded with transform skip, wherein the residual values are indicative of a difference between the block and a prediction block, and wherein, in transform skip, the residual values are not transformed from a sample domain to a frequency domain. The method includes determining a type of coding scheme to apply to the residual values based on the one or more syntax elements, determining the residual values based on the determined type of coding scheme, and reconstructing the block based on the determined residual values and the prediction block.

Signaling coding scheme for residual values in transform skip for video coding

An example method includes determining a color component of a unit of video data; determining, based at least on the color component, a context for context-adaptive binary arithmetic coding (CABAC) a syntax element that specifies a value of a low-frequency non-separable transform (LFNST) index for the unit of video data; CABAC decoding, based on the determined context and via a syntax structure for the unit of video data, the syntax element that specifies the value of the LFNST index for the unit of video data; and inverse-transforming, based on a transform indicated by the value of the LFNST index, transform coefficients of the unit of video data.

Context modeling for low-frequency non-separable transformation signaling for video coding

Growing community of inventors

Carlsbad, CA, United States of America

Muhammed Zeyd Coban