Prediction-Dependent Residual Scaling for Video Coding Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current design of Luma Mapping with Chroma Scaling (LMCS) in video coding standards like VVC faces issues with mapped precision of inter prediction samples exceeding the dynamic range of the internal coding depth, leading to inefficiencies in video compression.
Innovation Solution
Implementing Prediction Dependent Residual Scaling (PDRS) that adjusts the dynamic range of input signals using forward and inverse mapping functions, combined with luma-dependent chroma residual scaling, to optimize coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Luma Mapping with Chroma Scaling (LMCS) is used to compress video data, then compression efficiency is improved, but the mapped precision of inter prediction samples exceeds the dynamic range of the internal coding depth
Solution Approach 1:
The patent applies parameter changes by modifying the precision of prediction samples through scaling operations. Specifically, it scales the prediction samples to fit within the dynamic range of the internal coding depth, thereby resolving the overflow issue while maintaining compression efficiency. The scaling factor is derived based on the luma prediction samples and applied to adjust the chroma residual samples, ensuring they remain within valid precision bounds.
Solution Approach 2:
The patent introduces dynamic scaling factors that adapt to the local characteristics of the video content. The scaling factor is not fixed but is derived dynamically from the luma prediction samples, allowing the system to adjust the precision of chroma residuals according to the actual signal characteristics. This dynamic approach enables the system to handle varying precision requirements across different regions of the video frame.
2Measurement precision
If forward mapping scaling factors in 11-bit precision are used, then mapping accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial precision by using 11-bit precision for forward mapping scaling factors, which provides sufficient accuracy for the mapping operation without requiring the full precision that would be computationally expensive. This partial precision approach achieves the necessary mapping accuracy while keeping the computational complexity manageable by not using higher precision representations.
Data Source
AI summary
Methods for video encoding or decoding, devices, and non-transitory computer readable storage mediums are provided. A method for video encoding includes: dividing a picture into one or more coding units (CU); obtaining a plurality of prediction samples, in a mapped domain, of luma component of a current CU; obtaining a plurality of residual samples, in the mapped domain, of the luma component of the current CU; adding the plurality of prediction samples to the plurality of residual samples, resulting in a plurality of reconstructed samples, in the mapped domain, of the luma component of the current CU; converting the plurality of reconstructed samples from the mapped domain into an original domain based on a pre-defined plurality of inverse mapping scaling factors; and obtaining prediction information of the current CU based on the plurality of reconstructed samples in the original domain to form a video bitstream.