Variable-rate neural network based compression
AssigneesDouyin Vision (Beijing) Co., Ltd., Bytedance Inc.
InventorsZhaobin Zhang, Semih Esenlik, Yaojun Wu, Yue Li, Kai Zhang, Li Zhang
Life of this patent
Estimating the term…
Abstract
A mechanism for processing video data is disclosed. A determination is made to resize an image with a first size to create a resized image with a second size. A conversion is performed between a visual media data and a bitstream based on the resized image. The conversion includes applying a neural network-based coding model to the resized image to achieve variable rate neural network-based compression.
Claims20
A method for processing video data, comprising:
- determining to resize a first image with a first size to create a second image with a second size, wherein the first size is different from the second size;
- and
- performing a conversion between a visual media data and a bitstream based on the second image,
- wherein the conversion includes applying a neural network-based coding model to the second image to achieve variable rate neural network-based compression,
- wherein a set of candidate models used to code an image signal depends on whether a usage of the resizing of the first image occurs before or after a coding process;
- wherein one candidate model or multiple candidate models are determined depending on whether the usage of the resizing of the first image occurs before or after the coding process, or wherein at least one candidate model of the multiple candidate models is selected and signaled in the bitstream, or wherein at least one candidate model of the multiple candidate models is derived, or wherein the multiple candidate models correspond to multiple bitrates.
An apparatus for processing video data comprising a processor and a non-transitory memory with instructions thereon, wherein the instructions upon execution by the processor, cause the processor to:
- determine to resize a first image with a first size to create a second image with a second size, wherein the first size is different from the second size;
- and
- perform a conversion between a visual media data and a bitstream based on the second image,
- wherein the conversion includes applying a neural network-based coding model to the second image to achieve variable rate neural network-based compression,
- wherein a set of candidate models used to code an image signal depends on whether a usage of the resizing of the first image occurs before or after a coding process;
- wherein one candidate model or multiple candidate models are determined depending on whether the usage of the resizing of the first image occurs before or after the coding process, or wherein at least one candidate model of the multiple candidate models is selected and signaled in the bitstream, or wherein at least one candidate model of the multiple candidate models is derived, or wherein the multiple candidate models correspond to multiple bitrates.
A non-transitory computer-readable recording medium storing a bitstream of a video which is generated by a method performed by a video processing apparatus, wherein the method comprises:
- determining to resize a first image with a first size to create a second image with a second size, wherein the first size is different from the second size;
- and
- generating the bitstream based on the determining,
- wherein the generating includes applying a neural network-based coding model to the second image to achieve variable rate neural network-based compression,
- wherein a set of candidate models used to code an image signal depends on whether a usage of the resizing of the first image occurs before or after a coding process;
- wherein one candidate model or multiple candidate models are determined depending on whether the usage of the resizing of the first image occurs before or after the coding process, or wherein at least one candidate model of the multiple candidate models is selected and signaled in the bitstream, or wherein at least one candidate model of the multiple candidate models is derived, or wherein the multiple candidate models correspond to multiple bitrates.
Dependencies are inferred from the claim text.