Qualcomm Incorporated

In Suk Chong

Marta Karczewicz

Xianglin Wang

Cheng-Teh Hsieh

Wei-Jung Chien

Geert Van Der Auwera

Jaehong Chon

Chengjie Tu

Yang Yu

Sungwon Lee

Xin Zhao

Liwei Guo

Woo-Shik Kim

Sanjeev Kumar

Hariharan Ganesh Lalgudi

Jianle Chen

Vadim Seregin

Ying Hai Chen

Muhammed Zeyd Coban

Joel Sole Rojals

Athanasios Leontaris

Krishnakanth Rapaka

Wei Pu

Sumit Mohan

Prasanjit Panda

Siping Tao

Mang Li

Mayank Tiwari

Shashidhar Jangili

Ying Chen

A filter unit of a video encoder or video decoder can determine a first metric for a group of pixels within a block of pixels, determine a second metric for the group of pixels, determine a filter based on the first metric and the second metric, and generate a filtered image by applying the filter to the group of pixels. The first metric and second metric can be an activity metric and a direction metric, respectively, or can be other metrics such as an edge metric, horizontal activity metric, vertical activity metric, or diagonal activity metric.

Multi-metric filtering

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

Certain aspects of the present disclosure provide a method of encoding data. The method generally includes receiving data comprising a fractional number comprising an exponential component and a fractional component, the exponential component being represented by an exponential bit sequence, the fractional component being represented by a fractional bit sequence. The method further includes determining if the fractional component is within a threshold of 0 or 1. The method further includes setting the fractional component to 0 when the fractional component is within the threshold of 0 or 1. The method further includes downscaling the fractional bit sequence based on a difference between the exponential component and a second threshold. The method further includes encoding the data. The method further includes transmitting the encoded data.

Floating-point data compression

In general, techniques are described for performing multiple passes of sample adaptive offset (SAO) filtering when coding video data. A video decoding device comprising one or more processors may perform the techniques. The processors may determine a first SAO pixel classification for a block of video data and determine a first offset value based on the first SAO pixel classification and one or more pixel values of the block. The one or more processors may also determine a second SAO pixel classification for the block and determine a second offset value based on the second SAO pixel classification and the one or more pixel values of block of video data. The processors may then apply the first offset value and the second offset value to the block of video data to generate a block of SAO filtered video data.

Multi-type parallelized sample adaptive offset in video coding

This disclosure describes techniques for achieve high coding efficiency by periodically encoding anchor frames with a lower Quantization Parameter (QP) to provide better prediction for the following frames. Techniques include adaptively deciding which frames are encoded with lower QP by use of the collected encoding statistics via an encoding scheme with or without rate control.

Adaptive anchor frame and quantization parameter decision in video coding

A video coding device may be configured to estimate, based on a combination of a first parameter and a number of non-zero coefficients in a frame, a number of bits for non-zero coefficients of the frame, to encode the frame based on the estimated number of bits for the non-zero coefficients, to collect an actual number of bits used to encode the non-zero coefficients of the frame and an actual number of the non-zero coefficients in the frame, to update, based on the actual number of bits used to encode the non-zero coefficients of the frame and the actual number of non-zero coefficients in the frame, only the first parameter to form an updated first parameter, to form a rate estimation model using the updated first parameter and a second parameter, and to select, based on the rate estimation model, a coding mode for each block in the frame.

Zero block detection using adaptive rate model

Systems and methods for encoding and decoding video data are disclosed. The method can include signaling in syntax information a picture parameter set (PPS) indicating a first tile size partition. The method can also include storing a plurality of tile size partitions and associated PPS identifiers (PPSID) in a database. If a second tile size partition for a second frame of video data is the same as a tile size partition stored in the database, the method can include signaling the PPSID for the corresponding tile size partition. If the second tile size partition is not the same as a tile size partition stored in the database, the method can include signaling a new PPS with the second tile size partition. The system can provide an encoder and a decoder for processing the video data encoded by the method for encoding video data.

Method and apparatus for video coding using adaptive tile sizes

In one example, a device for accessing image data includes a memory configured to store image data, the memory comprising a first region and a second region; and one or more processing units configured to code most significant bits (MSBs) of a plurality of residuals of samples of a block of an image, each of the residuals representing a respective difference value between a respective raw sample value and a respective predicted value for the respective raw sample value, access the coded MSBs in the first region of the memory, and access least significant bits (LSBs) of the plurality of residuals of the samples in the second region of the memory.

Storing and retrieving high bit depth image data

A device for encoding video data may be configured to encode video data according to a set of sample adaptive offset (SAO) types; perform a plurality of coding passes to test a subset of the SAO types for a first block of video data, wherein the subset is smaller than the set; select from the subset of SAO types an SAO type for the first block of video data; and generate for inclusion in an encoded bitstream, information for identifying the selected SAO type for the first block.

Low complexity sample adaptive offset (SAO) coding

In an example, aspects of this disclosure relate to a method of coding video data that includes identifying a plurality of quantization parameter (QP) values associated with a plurality of reference blocks of video data. The method also includes generating a reference QP for the plurality of reference blocks based on the plurality of QPs. The method also includes storing the reference QP, and coding a block of video data based on the stored reference QP.

Quantization in video coding

In one example, a device for accessing image data includes a memory configured to store image data and one or more processors configured to code a plurality of bit length values for a plurality of block fixed length code length (bflc_len) values for a plurality of blocks of a tile or sub-tile of an image, the bit length values representing numbers of bits used to code the blfc_len values, code the bflc_len values for each of the plurality of blocks such that the bflc_len values have numbers of bits indicated by the respective bit length values, code the codewords for each of the plurality of blocks such that the codewords have the numbers of bits indicated by the bflc_len values for corresponding blocks of the plurality of blocks, and access the bit length values, the bflc_len values, and the codewords in the memory.

A video encoding device comprises a memory configured and at least one processor configured to: determine whether a metric meets a condition based on statistics, wherein the statistics are associated with a first video encoding mode checking order and a second video encoding mode checking order, responsive to determining that the metric meets the condition, select a first encoding mode checking order to encode the first block of video data responsive to determining that the condition is not met, select a second encoding mode checking order different from the first encoding mode checking order to encode the first block of video data, update the statistics based on the selected first or second encoding mode checking order, and encode a second block of video data, based on the updated statistics, and using the first or second mode checking order.

Adaptive mode checking order for video encoding

A video coder configured to perform sample adaptive offset filtering can determine a center value for a set of pixels based on values of pixels in the set, divide bands of pixels values into groups based on the center value, and determine offset values for the bands based on the groups.

Adaptive center band offset filter for video coding

To encode video data, a video encoder partitions a 2N&#xd7;2N block of video data into four N&#xd7;N blocks, determines encoding modes for each of the four N&#xd7;N blocks, calculates values representative of encoded versions of the four N&#xd7;N blocks using the encoding modes for each of the four N&#xd7;N blocks, determines whether to skip testing of at least one non-square partitioning mode for the 2N&#xd7;2N block based on the calculated values, and encodes the 2N&#xd7;2N block based at least in part on the determination of whether to skip testing of the at least one non-square partitioning mode.

Growing community of inventors

San Diego, CA, United States of America

In Suk Chong