Predictive Picture Encoding With Mapped Residual Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face inefficiencies when adapting to new color gamut and dynamic range standards, leading to information loss and reduced coding performance due to fixed-point precision mapping and inverse mapping processes.
Innovation Solution
Adapting the quantization step based on a mapping function to modify the quantization of transformed prediction residuals, optimizing bitstream efficiency and reconstruction quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If mapping/reshaping of samples is performed before encoding to fit new container properties, then sample distribution compatibility is improved, but information loss increases due to fixed-point precision mapping and inverse mapping
Solution Approach 1:
The patent applies mapping/reshaping of samples before encoding to pre-adjust sample distribution to match the expected properties of the target container (e.g., BT.2100). This preliminary action ensures compatibility with new color gamut and dynamic range standards, allowing current codecs to process wide-gamut content effectively. The mapping is performed in advance using fixed-point precision arithmetic to maintain efficiency while adapting to new container requirements.
Solution Approach 2:
The patent applies inverse mapping after decoding to reverse the initial mapping transformation. This inversion process restores the decoded samples to their original wide-gamut representation, compensating for the distortion introduced by the preliminary mapping. By applying the inverse operation, the system recovers information that would otherwise be permanently lost, reducing overall information loss while maintaining compatibility throughout the encoding-decoding chain.
2Productivity
If quantization step is adapted per block based on prediction residual coefficients, then local coding efficiency is improved, but overall coding performance deteriorates when blocks contain samples with many different values
Solution Approach 1:
The patent adapts the quantization step size locally for each block based on the statistical properties of the prediction residual coefficients within that block. By analyzing the energy distribution and variance of residuals in each block, the system applies finer quantization to blocks with high variability (containing diverse sample values) and coarser quantization to blocks with low variability. This local adaptation optimizes the trade-off between bit rate and reconstruction quality for each block, improving overall coding performance while maintaining efficiency.
Data Source
AI summary
A method for encoding a block of a picture comprising, for at least one sample of the block and for one current component: obtaining a prediction value; determining a mapped residual value from a source value of the sample and from the prediction value responsive to a mapping function; and encoding the mapped residual value into a bitstream; wherein the mapping function is derived to obtain at least one of a reduction of a bit cost of the bitstream for a given quality of reconstruction or an increase of quality of reconstruction for a given bit cost of the bitstream.