Parallel Video Coding via Diagonal Block Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The HEVC coding standard faces challenges in achieving ideal parallelism and coding quality due to large search spaces and data dependencies in intra-frame mode decision, particularly in reconstructed pixel, most probable mode computing, and probabilistic characteristic model dependencies, which hinder parallel processing of video coding.
Innovation Solution
The method involves dividing a video image into pixel blocks and performing parallel coding on blocks along diagonal lines, where each block references coded blocks from adjacent diagonal lines, thereby eliminating dependencies and allowing for efficient parallel processing across CPU and GPU architectures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If intra-frame mode decision is performed using traditional sequential processing, then coding quality is maintained, but coding speed is limited due to data dependencies
Solution Approach 1:
The video image is divided into multiple CTUs (coding tree units), and each CTU is further divided into CUs (coding units). This segmentation allows different CTUs to be processed in parallel while maintaining intra-frame prediction quality within each CTU by using reconstructed pixels from previously coded CTUs.
Solution Approach 2:
The patent introduces a new processing dimension by organizing CTUs in raster scan order and enabling parallel processing across different CTUs while maintaining sequential dependency only within the reconstructed pixel references. This transforms the problem from a fully sequential process to a partially parallelizable process across spatial dimensions.
2Productivity
If parallel processing is implemented using Tiles or WPP algorithms, then coding speed is improved, but ideal balance between parallelism degree and coding quality is not achieved
Solution Approach 1:
The patent applies different processing strategies to different regions: within each CTU, sequential processing ensures quality by maintaining proper dependency relationships for intra-frame prediction, while across different CTUs, parallel processing improves productivity. This local differentiation of processing quality ensures overall coding quality while achieving parallelism.
3Loss of time
If search space is reduced to improve coding speed, then parallel processing becomes easier, but rate-distortion optimization is compromised
Solution Approach 1:
The patent performs preliminary actions by dividing the image into CTUs and establishing the parallel processing framework before actual mode decision. This allows mode decision to be performed in parallel across multiple CTUs without reducing the search space, thereby maintaining rate-distortion optimization while reducing total mode decision time through parallel execution.
Data Source
AI summary
Aspects of the present disclosure provide a method and an apparatus for video coding. In some examples, the apparatus includes processing circuitry that divides a first video image into a plurality of pixel blocks of the same size. The processing circuitry selects, from the plurality of pixel blocks of the first video image, one or more pixel blocks on a straight line that is parallel to or overlapping with a diagonal line of the first video image, one or more pixel blocks on a neighboring parallel line of the straight line being previously coded. Then the processing circuitry performs a parallel coding on the one or more pixel blocks on the straight line based on the one or more coded pixel block on the neighboring parallel line.