Video Coding Block Vector Prediction for Spatial Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques fail to effectively utilize spatial similarity between blocks in the same picture for prediction, limiting coding efficiency, especially when similar patterns recur in a picture.
Innovation Solution
A method and apparatus for video encoding/decoding that derive a block vector predictor from a reference block and generate a prediction block using this predictor, allowing for spatial prediction between blocks in the same picture, and efficiently signal the selected prediction vector through a candidate list and predictor index.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If intra-prediction using neighboring blocks is used, then spatial compression is achieved, but blocks spatially apart from each other cannot be utilized for prediction
Solution Approach 1:
The patent extends the prediction dimension from local neighboring blocks to distant blocks within the same picture by introducing block vectors that can point to any location in the current picture, enabling long-range spatial prediction while maintaining the intra-prediction framework
2Productivity
If inter-prediction is used, then temporal similarity is exploited, but regions in the current picture cannot be used as prediction values
Solution Approach 1:
The patent inverts the traditional inter-prediction approach by using blocks within the current picture as reference instead of using previous or future pictures, thereby enabling current-picture-based prediction that was previously unavailable in inter-prediction frameworks
3Productivity
If similar patterns occur repeatedly in a picture, then prediction on the pattern can increase coding efficiency, but existing techniques cannot effectively utilize spatial similarity between blocks
Solution Approach 1:
The patent implements a copying mechanism where block vectors reference and copy pixel data from distant blocks within the same picture, enabling efficient replication of recurring patterns while maintaining the ability to handle diverse spatial configurations
Data Source
AI summary
A method for decoding an image, according to the present invention, can comprise the steps of: drawing out a vector predictor of a block vector for indicating a reference block with respect to a current block to be predicted within a current picture; drawing out the block vector on the basis of the vector predictor and a vector differential value corresponding to a different value of the block vector and the vector predictor; and generating a prediction block with respect to the current block on the basis of the block vector.