Neural-Network Post-Processing Filters for 4:0:0 Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies, such as VVC and VSEI, face issues with neural-network post-filter characteristics that do not support applications with 4:0:0 input, bit-depth increase, frame rate change, and missing or incorrect signaling of syntax elements, leading to inefficiencies and limitations in video processing.
Innovation Solution
Specify new purposes and syntax elements for neural-network post-processing filters (NNPF) to handle 4:0:0 input, bit-depth increase, and frame rate change, including signaling chroma format changes, bit-depth differences, and ensuring proper derivation of output tensors, while ensuring mutual exclusivity of colorization and chroma upsampling purposes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If neural-network post-processing filter is applied with comprehensive chroma input indications, then filtering accuracy is improved, but coding efficiency deteriorates due to redundant indications in 4:0:0 color format
Solution Approach 1:
The patent applies local quality by making the chroma input tensor indication conditional on the actual color format of the input picture. When the input is in 4:0:0 format without chroma component, the chroma input indication is omitted or set to a default value. When chroma component exists, the indication is provided with full detail. This localized adaptation of information provision resolves the contradiction between comprehensive filtering accuracy and coding efficiency.
2Adaptability or versatility
If new syntax elements are added to support 4:0:0 input and bit-depth increase, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent implements universality by designing the NNPF purpose field to accommodate multiple functions: colorization, chroma upsampling, bit-depth increase, and frame rate change. A single unified syntax structure with purpose-specific conditions handles diverse applications. This multi-functional design improves adaptability while controlling device complexity by avoiding separate dedicated syntax elements for each application type.
Solution Approach 2:
The patent applies dynamics by making the presence and interpretation of syntax elements conditional on the NNPF purpose value. Different purpose values (colorization, chroma upsampling, bit-depth increase, frame rate change) activate different syntax element requirements and derivation rules. This dynamic syntax structure adapts to the specific application, providing necessary complexity only when needed while maintaining simplicity for other cases.
3Manufacturing precision
If chroma format change indication is always signaled, then output color accuracy is improved, but bitstream size increases for cases where chroma format remains unchanged
Solution Approach 1:
The patent applies discarding and recovering by selectively signaling chroma format change indication only when the output chroma format differs from the input chroma format. When the format remains unchanged, the indication is discarded (omitted or set to default). The decoder recovers the original format information from the input picture parameters. This selective signaling reduces bitstream size while maintaining output color accuracy when format changes occur.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. The method comprises: performing a conversion between a video and a bitstream of the video, wherein a neural-network post-processing filter (NNPF) is applied on at least one picture associated with the video, and if a color format of the at least one picture is a first color format without a chroma component, at least one indication associated with a chroma input tensor for the NNPF is absent from the bitstream.