Multiple Reference Line Intra Prediction for Wide-Angle Pixel Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like HEVC and VVC face challenges in accurately predicting pixel blocks due to limitations in intra-prediction methods, particularly for non-square blocks with wide-angular directions and non-integer sample positions, leading to inefficiencies in compression and reconstruction.
Innovation Solution
The method involves using multiple reference lines for intra-prediction by blending samples from adjacent and non-adjacent lines, applying interpolation filters when necessary, and utilizing explicit or implicit indications to determine the use of these lines based on block characteristics, thereby enhancing prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intra-prediction methods are used for non-square blocks with wide-angular directions, then the coding process is simple, but the prediction accuracy deteriorates
Solution Approach 1:
The reference samples are segmented into multiple reference lines based on the intra-prediction angle. For wide-angular directions, samples are divided between a first reference line and a second reference line, allowing each line to contribute appropriate samples to the prediction process. This segmentation enables accurate prediction for non-square blocks by adapting to the specific angular direction.
Solution Approach 2:
The patent extends the traditional single reference line approach by introducing multiple reference lines in different spatial dimensions. Instead of relying on a single horizontal or vertical reference line, the method utilizes both horizontal and vertical reference lines (first and second reference lines) to capture wide-angular directions more accurately, adding a dimensional aspect to the reference sampling process.
2Measurement precision
If multiple reference lines are used for intra-prediction, then the prediction accuracy for non-square blocks improves, but the computational complexity increases
Solution Approach 1:
The method applies multiple reference lines selectively rather than universally. It uses the first reference line for standard directions and only invokes the second reference line when wide-angular directions are detected. This partial application of the complex method maintains accuracy where needed while avoiding unnecessary computational overhead in simpler cases.
Solution Approach 2:
The patent dynamically changes the prediction parameters (reference line selection, sample positioning) based on the intra-prediction angle and block characteristics. By adjusting which reference lines to use and how to position samples based on the specific angle, the method adapts to different scenarios, improving accuracy for wide-angular directions while maintaining efficiency for standard cases.
3Measurement precision
If reference samples are taken from non-adjacent lines, then the prediction accuracy for wide-angular directions improves, but the complexity of determining sample positions increases
Solution Approach 1:
The patent pre-establishes the relationship between intra-prediction angles and reference line selections. The decoder is预先 configured to understand which angles correspond to which reference line combinations, eliminating the need for complex real-time calculations. This preliminary setup simplifies the sample position determination process while maintaining accuracy.
Solution Approach 2:
The patent introduces an intermediary mechanism (the second reference line) that mediates between the prediction requirement and the sample source. Instead of directly calculating complex sample positions from non-adjacent lines, the second reference line serves as an intermediary that provides pre-organized samples suitable for wide-angular directions, simplifying the overall process.
Data Source
AI summary
A method for using multiple reference lines to perform intra-prediction of a block of pixels is provided. A video coder receives data for a block of pixels to be encoded or decoded as a current block of a current picture of a video. The video coder receives or signals a selection of first and second reference lines among a plurality of reference lines that neighbor the current block. The video coder blends samples of first and second reference lines, each blended sample generated by combining a sample in the first reference line with a sample in the second reference line that is identified according to an intra-prediction mode. The video coder generates a prediction of the current block with the intra-prediction mode by using the blended samples. The video coder encodes or decodes the current block by using the generated prediction.