Intra Prediction Using Multi-Resolution Reference Pixels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image encoding methods, such as those in the HEVC standard, face reduced coding efficiency due to low correlation between co-located blocks in InterLayer images, leading to suboptimal prediction and increased encoding errors.
Innovation Solution
The proposed method generates a second decoded image with higher resolution or quality by using pixels from the first decoded image as reference for intra prediction, specifically employing pixels from the second decoded image as top and left references and those from the first decoded image as right and bottom references for the current block.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional intra prediction using only top and left reference pixels is used, then the encoding process is simple, but prediction accuracy deteriorates when correlation between top/left and current pixels is low
Solution Approach 1:
The patent extends the traditional intra prediction by utilizing reference pixels from multiple decoded images (first decoded image and second decoded image) in addition to the current decoded image. This multi-dimensional approach allows the system to select the most correlated reference pixels from different image sources, significantly improving prediction accuracy when traditional top/left references have low correlation with the current block
2Measurement precision
If only the second decoded image with higher resolution is used for prediction, then prediction quality improves, but encoding efficiency decreases due to increased bitstream size
Solution Approach 1:
The patent applies different reference pixel strategies to different blocks based on their specific characteristics. For each current block, the system evaluates correlation between the block and reference pixels from both the first decoded image and second decoded image, then selectively uses the most appropriate reference source. This local optimization ensures high prediction quality where needed while maintaining encoding efficiency for blocks where simpler methods suffice
Solution Approach 2:
The system dynamically changes the reference pixel parameters by selecting from multiple decoded images with different resolutions and qualities. Instead of using a fixed reference image, the patent adapts the reference selection based on correlation metrics, allowing the encoder to optimize the balance between prediction quality and bitstream size for each block
Data Source
AI summary
An image encoding method includes: generating a first bitstream by encoding an input image, and a first decoded image by encoding and decoding the input image; and generating a second bitstream by encoding the input image using the first decoded image, and a second decoded image by encoding and decoding the input image; wherein the second decoded image has resolution or image quality higher than resolution or image quality of the first decoded image, the generating of a second bitstream and a second decoded image includes generating a prediction block for a current block by intra prediction, and in the generating of a prediction block, the prediction block is generated using (i) pixels included in the second decoded image as top and left reference pixels for the current block, and (ii) pixels included in the first decoded image as right and bottom reference pixels for the current block.