Image Coding Apparatus Dynamic Tile Division for Motion Vector Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In image coding methods like HEVC, motion vector detection accuracy is compromised when the search range is too small or too large, leading to increased circuit scale and processing complexity, particularly due to the need for large line buffers that consume more resources as image resolution increases.
Innovation Solution
The technique involves dividing the target image into regions in the horizontal direction and adjusting the number of tile divisions based on the temporal distance between the target and reference images, allowing for a dynamic search range that matches the temporal distance while minimizing the capacity of the line buffer and reducing bus bandwidth consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the search range is enlarged to improve motion vector detection accuracy for long temporal distances, then the accuracy is improved, but the line buffer capacity and circuit scale must be increased
Solution Approach 1:
The target image is divided into multiple tiles in the horizontal direction, allowing the encoding to be performed on smaller regions independently. This segmentation enables the line buffer to store only the reference image portions needed for each tile's search range, rather than requiring a large buffer for the entire image, thus reducing the line buffer capacity while maintaining adequate search range for motion vector detection
Solution Approach 2:
Different search ranges are applied to different tiles based on their local requirements. The line buffer stores reference image data with dimensions tailored to each tile's search needs rather than uniformly allocating buffer space for the maximum possible search range across the entire image, optimizing the balance between search accuracy and buffer capacity
2Reliability
If the line buffer capacity is increased to support larger search ranges, then the motion tracking accuracy is improved, but the circuit scale increases
Solution Approach 1:
By dividing the encoding process into multiple tiles processed sequentially, the circuit only needs to support the search range requirements for individual tiles at any given time. This reduces the peak memory capacity and associated circuit complexity compared to supporting a full-image search range simultaneously, while still achieving reliable motion tracking within each tile's context
3Quantity of substance
If the search range is reduced to decrease circuit scale, then the line buffer capacity is reduced, but the motion vector detection accuracy deteriorates
Solution Approach 1:
The tile division strategy dynamically adapts the effective search range to match the temporal distance between the target image and reference image. When temporal distance is large, more tiles are used to provide adequate search coverage; when temporal distance is small, fewer tiles are sufficient. This dynamic approach ensures adequate motion vector detection accuracy while minimizing line buffer capacity requirements for each specific encoding scenario
Data Source
AI summary
A detection unit successively selects each of a plurality of blocks of a target image as a target block and detects a motion vector of the target block using a reference image. During motion vector detection, a control unit performs control to partially store the reference image into a storage unit so as to achieve a state where a search range for the target block is stored. The detection unit selects each of the plurality of blocks as the target block in an order of raster scan for each of N regions obtained by dividing the target image in a horizontal direction. When a temporal distance between the target image and the reference image is a second distance that is longer than a first distance, a value of N is large compared to a case where the temporal distance is the first distance.