Unified Depth Intra Prediction Using Extended Neighboring Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional video coding systems, specifically in 3D-HEVC, face complexity in depth map Intra prediction due to multiple partition patterns, which complicates the depth Intra coding process without significant performance simplification.
Innovation Solution
The method simplifies depth map coding by unifying conventional case-1 and case-2 DMM prediction modes into a unified DMM prediction mode, using extended reconstructed neighboring depth samples to derive prediction values for segment partitioning, reducing the number of cases from 4 to 3 and incorporating extended top or left neighboring samples for improved prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple partition patterns (4 cases) are used for DMM prediction, then prediction accuracy is improved, but device complexity and processing difficulty increase
Solution Approach 1:
The patent merges case-1 and case-2 into a unified prediction mode, and merges case-3 and case-4 into another unified mode. This reduces the number of separate cases from 4 to 2, simplifying the processing logic while maintaining prediction accuracy through the use of extended neighboring samples.
Solution Approach 2:
The unified prediction mode serves multiple functions by handling both case-1 and case-2 scenarios with the same processing logic. The extended neighboring samples (including virtual samples at block boundaries) provide a universal approach that works for different partition patterns without requiring separate handling for each case.
2Reliability
If conventional DMM prediction with 4 cases is used, then prediction performance is maintained, but encoding and decoding time increase
Solution Approach 1:
By combining case-1 and case-2 into a unified mode, and case-3 and case-4 into another unified mode, the patent reduces the number of conditional checks and processing steps required during encoding and decoding, thereby reducing time loss while maintaining prediction performance through the use of extended neighboring samples.
3Measurement precision
If extended neighboring samples are used for prediction, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary action by extending the neighboring samples (creating virtual samples at block boundaries) before the actual prediction process. This preparation step is done once and reused across different prediction modes, improving accuracy without proportionally increasing complexity during the main prediction operation.
Data Source
AI summary
A method of depth map coding for a three-dimensional video coding system incorporating extended reconstructed neighboring depth samples is disclosed. The depth block is partitioned into one or more segments. A prediction value for each segment is derived based on reconstructed neighboring depth samples. The reconstructed neighboring depth samples for at least one segment include at least one reconstructed sample from an extended top neighboring row or an extended left neighboring column. The extended top neighboring row includes at least one extended top-row reconstructed depth sample located adjacent to top side of a second depth block adjacent to right side of the current depth block. The extended left neighboring column includes at least one extended left-column reconstructed depth sample located adjacent to left side of a third depth block adjacent to bottom side of the current depth block.