Region-Based Inter/Intra Prediction for 360-Degree Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compression methods for 360-degree videos do not effectively utilize spatial correlations between different viewpoints, leading to inefficient encoding and decoding processes.
Innovation Solution
A region-based inter/intra prediction technique is employed to determine spatial correlation degrees between blocks in 360-degree videos, constructing merge candidate lists that include blocks from opposite boundary faces and excluding temporally or spatially neighboring blocks when necessary, using projection formats like cube map projection (CMP) and equi-rectangular projection (ERP).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional inter prediction technique is used for 360-degree video, then encoding process is simple, but spatial correlations between different viewpoints are not effectively utilized leading to poor compression efficiency
Solution Approach 1:
The patent segments the 360-degree video into multiple regions based on projection format (cube map faces or ERP zones). Each region is independently processed with region-specific merge candidate lists, allowing spatial correlations to be exploited within each segment while maintaining manageable complexity per region.
Solution Approach 2:
The patent introduces a new dimension for selecting merge candidates by considering blocks from opposite boundary faces in the projection space. For cube map projection, this means selecting candidates from diagonally opposite faces; for ERP, this means selecting from corresponding zones in the unfolded sphere. This dimensional approach captures spatial correlations that conventional temporal-only prediction misses.
2Productivity
If region-based prediction with opposite boundary face blocks is used, then spatial correlations are leveraged improving compression efficiency, but merge candidate list construction becomes more complex
Solution Approach 1:
The patent applies different merge candidate selection strategies to different regions based on their projection format. Cube map faces use diagonal opposite face candidates, while ERP zones use corresponding zone candidates. This local customization optimizes compression for each region's specific geometric properties without imposing unnecessary complexity on all regions.
Solution Approach 2:
The patent performs preliminary determination of the projection format and identification of opposite boundary faces/zones before constructing merge candidate lists. This preliminary action pre-establishes the geometric relationships and candidate selection rules, simplifying the actual merge candidate list construction process during encoding.
3Productivity
If temporal neighboring blocks are excluded from candidate blocks, then region-based spatial prediction is enhanced, but the number of available candidate blocks decreases
Solution Approach 1:
The patent compensates for excluding temporal neighbors by introducing spatial neighbors from opposite boundary faces/zones as a new dimension of candidate blocks. This spatial dimension provides alternative correlated blocks that maintain or improve prediction accuracy while enabling region-based prediction strategies.
Data Source
AI summary
Disclosed are region-based image encoding/decoding method and apparatus. The method of decoding an image includes: determining a projection format of a 360-degree video image; determining, on the basis of the determined projection format, a spatial correlation degree between a current block and a neighboring block; and determining, on the basis of the determined spatial correlation degree, a candidate block for predicting the current block.