Neural-Network Post-Processing for Video Bit-Depth Increase
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies do not adequately support applications requiring bit depth increase, such as converting from standard dynamic range (SDR) to high-dynamic range (HDR), and lack proper signaling for neural-network post-processing filters (NNPF) in video bitstreams, leading to inefficiencies and limitations in video processing.
Innovation Solution
Implement a neural-network post-processing filter (NNPF) that supports bit depth increase and provides enhanced signaling for various video processing applications, including bit depth upsampling, frame rate changes, and chroma format adjustments, through specific syntax element definitions and SEI message enhancements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional video coding techniques are used, then coding simplicity is maintained, but coding efficiency and support for advanced applications (such as bit depth increase) are insufficient
Solution Approach 1:
The neural-network post-processing filter (NNPF) is designed to perform multiple functions including bit depth increase, frame rate changes, and chroma format adjustments through a single unified filter architecture. This allows the system to handle diverse video processing applications without requiring separate dedicated tools for each function, thereby improving adaptability while managing complexity.
Solution Approach 2:
The patent utilizes parameter changes in the neural network configuration to adapt the NNPF for different processing tasks. By modifying network parameters such as input/output bit depths, temporal downsampling factors, and chroma subsampling configurations, the same filter architecture can be flexibly applied to various video coding scenarios including SDR to HDR conversion and different chroma formats.
2Manufacturing precision
If NNPF is applied for bit depth increase, then coding quality is improved, but device complexity and signaling requirements increase
Solution Approach 1:
The patent extracts and separates the bit depth increase function into a dedicated neural-network post-processing filter stage, distinct from the main video coding process. This allows the complex bit depth transformation to be handled by a specialized component (NNPF) with its own optimized architecture, improving coding quality while containing the complexity within a modular unit rather than throughout the entire system.
Solution Approach 2:
The NNPF performs bit depth increase and other transformations as a preliminary post-processing step before final video output. By preparing the video data with appropriate bit depth and format conversions in advance, the system ensures higher coding quality in subsequent processing stages without requiring complex real-time adjustments during encoding or decoding.
3Adaptability or versatility
If existing NNPF signaling is used, then bitstream simplicity is maintained, but functionality and adaptability are limited
Solution Approach 1:
The signaling for NNPF is segmented into distinct parameters including separate indicators for bit depth increase, frame rate changes, and chroma format adjustments. This segmentation allows each function to be independently controlled and configured in the bitstream, enhancing adaptability while organizing signaling information in a structured manner that minimizes redundancy and information loss.
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, the bitstream comprises a first indication indicating a purpose of the NNPF, and one of candidates for the purpose is increasing a bit depth of a sample value in the at least one picture.