Reference Pixel Selection for Depth Map Intra Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques, particularly in 3D-HEVC, face challenges in improving intra prediction accuracy for depth maps, leading to visible coding artifacts at sharp edges and inefficiencies in encoding and decoding large video frames.
Innovation Solution
The proposed solution involves advanced reference sample selection and filtering techniques for intra prediction, where reference samples are chosen based on partition patterns and absolute differences, allowing for improved prediction of depth map partitions, and the use of extended neighboring blocks for more accurate DC prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional intra prediction techniques are used for depth maps, then encoding complexity is reduced, but coding accuracy deteriorates and visible artifacts appear at sharp edges
Solution Approach 1:
The current block is divided into two partitions (partition 0 and partition 1) based on partition patterns. Different reference samples are selected for each partition, allowing tailored prediction strategies for different regions. This segmentation enables more accurate prediction at sharp edges while maintaining manageable encoding complexity through structured partitioning.
Solution Approach 2:
Different reference sample selection strategies are applied to different partitions based on their local characteristics. Partition 0 uses reference samples from specific neighboring blocks, while partition 1 uses different reference samples, allowing each region to be predicted with quality optimized for its local structure and edge characteristics.
2Measurement precision
If reference samples from all neighboring blocks are used, then prediction accuracy improves, but the complexity of selecting and processing reference samples increases
Solution Approach 1:
The set of neighboring blocks is segmented into specific groups (upper-right neighboring block and lower-left neighboring block) for reference sample selection. This segmentation reduces the number of reference samples that need to be processed while maintaining prediction accuracy by focusing on the most relevant neighboring blocks for each partition.
3Productivity
If simple DC prediction is used, then encoding speed is maintained, but coding efficiency deteriorates and artifacts are visible
Solution Approach 1:
Different DC prediction approaches are applied to different partitions based on their local characteristics. Partition 0 and partition 1 use different reference samples for DC prediction, allowing each partition to have prediction quality optimized for its local structure, reducing artifacts at sharp edges while maintaining overall coding efficiency.
Data Source
AI summary
A video codec configured to receive a current block and a plurality of neighboring pixels, wherein the current block comprises a first partition and a second partition, select one or more reference pixels from the plurality of neighboring pixels, and predict a plurality of pixels located in the second partition based on the reference pixels.