HDR Video Encoding Using LDR Segmentation and Global Illumination Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for representing and encoding High Dynamic Range (HDR) content using Low Dynamic Range (LDR) content and HDR residuals are not optimal, leading to instability in prediction and potential artifacts like halos on high gradients, and lack efficiency in existing encoding schemes.
Innovation Solution
Encoding HDR videos by using LDR videos as references, extracting a further LDR video and global illumination data from the HDR video, and employing predictive encoding with lossless encoding of global illumination data to stabilize prediction and avoid artifacts, while allowing subsampling without introducing halos.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If HDR content is represented using LDR content and HDR residuals, then encoding efficiency is improved, but prediction stability deteriorates and artifacts like halos appear on high gradients
Solution Approach 1:
The patent segments the HDR video encoding into two independent LDR video encodings: one for the tone-mapped HDR content and another for the illumination difference content. This segmentation allows each LDR encoder to operate independently with stable predictions, avoiding the instability that would occur when directly encoding HDR residuals against LDR references with different bit depths and luminance ranges.
Solution Approach 2:
The patent introduces an intermediary representation by converting the HDR video into two separate LDR videos through tone mapping operations. These LDR intermediaries serve as stable references for predictive encoding, mediating between the HDR source content and the LDR encoding system, thereby preventing direct comparison issues between incompatible luminance ranges.
2Quantity of substance
If HDR content is represented using LDR content and HDR residuals, then bit depth reduction is achieved, but manufacturing precision deteriorates due to loss of luminance accuracy
Solution Approach 1:
The patent transforms the single-dimension HDR luminance representation into a two-dimensional LDR representation space by separating illumination information from reflected light information. This dimensional transformation allows the system to encode HDR content using two LDR videos, effectively trading one dimension of luminance precision for another dimension of spatial and temporal redundancy that can be efficiently compressed.
3Device complexity
If subsampling is applied to HDR content, then encoding complexity is reduced, but artifacts like halos appear on high gradients
Solution Approach 1:
The patent segments the HDR video into two LDR videos representing different physical quantities (reflected light and illumination). This segmentation allows independent subsampling of each component without the halo artifacts that occur when subsampling the full HDR signal, because each LDR component has a narrower luminance range that is more tolerant to subsampling operations.
Data Source
AI summary
The inventors realized that representing HDR content using LDR content and HDR residual is not unproblematic. The invention therefore proposes representing HDR content using LDR content, LDR residual and global illumination data, instead.That is, it is proposed a method of encoding an HDR video of high dynamic range together with an LDR video, the LDR video providing a lower dynamic range depiction of the HDR video content, the method comprising using processing means for encoding one video of the LDR video and a further LDR video extracted from the HDR video independent from the other video of the LDR video and the HDR video and predictive encoding the other video using the one video as reference, and lossless encoding global illumination data further extracted from the HDR video.This has the advantage that prediction is more stable and globally optimal predictors are more easily found.