Intra Prediction Reference Area Using In-Loop Filtered Pixels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as VVC and ECM, utilize non-final reconstructed pixels for intra prediction, leading to less accurate prediction due to the use of coded areas before in-loop filtering, which can result in inefficiencies and suboptimal coding performance.
Innovation Solution
Intra prediction is performed based on final reconstructed pixels of the coded area after in-loop filtering, utilizing the fully filtered reference area for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If non-final reconstructed pixels (coded area before in-loop filtering) are used for intra prediction, then the coding process can be simplified and executed faster, but the prediction accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by performing in-loop filtering on the reference area before using it for intra prediction. Specifically, the decoded picture buffer performs in-loop filtering (deblocking, SAO, ALF) on previously coded blocks, and these filtered pixels are then used as reference for current block prediction. This preliminary filtering action improves prediction accuracy without significantly impacting coding speed.
Solution Approach 2:
The decoded picture buffer acts as an intermediary between the in-loop filtering process and the intra prediction process. It stores both the non-filtered reconstructed pixels and the in-loop filtered pixels, allowing the intra prediction to selectively use the filtered version when available. This intermediary structure enables the system to benefit from filtering improvements while maintaining the original coding structure.
2Measurement precision
If in-loop filtering is applied to the reference area, then prediction accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent merges the in-loop filtering process with the existing decoded picture buffer structure. The same decoded picture buffer that stores reconstructed pixels is extended to also store in-loop filtered pixels. This merging approach allows the system to utilize filtering results without adding separate complex storage structures, thereby improving prediction accuracy while limiting the increase in processing complexity.
Solution Approach 2:
The decoded picture buffer is given multi-functionality by making it serve both as a storage for reconstructed pixels and as a buffer for in-loop filtered pixels. This universal structure allows the buffer to support both the traditional intra prediction using non-filtered pixels and the enhanced intra prediction using filtered pixels, without requiring dedicated separate structures for each function.
3Productivity
If in-loop filtered pixels are used for intra prediction, then coding efficiency is improved, but memory requirements increase
Solution Approach 1:
The patent adds a temporal dimension to the decoded picture buffer by organizing it to store both non-filtered and filtered pixels across different time states. The buffer maintains historical information in a structured manner where filtered pixels from previously coded blocks are preserved and made available for future prediction. This dimensional organization allows efficient memory utilization while providing access to filtered reference data.
Data Source
AI summary
An apparatus may be configured to: select a coding unit of an image; select a first area within a coded area of the image, wherein the first area comprises, at least, a first part and a second part, wherein the first part is at least partially different from the second part; perform in-loop filtering of, at least, the first part of the first area; perform intra prediction for the selected coding unit based, at least partially, on the first area, comprising the in-loop filtered first part; and output an intra-prediction block for the selected coding unit based on the performed intra prediction.