Template-Based Video Prediction With Selective Above or Left Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional template-based coding tools in video compression often fail to effectively capture the statistical properties of blocks due to assumptions about L-shaped templates being distorted by edges, particularly in non-camera captured content like gaming or screen content.
Innovation Solution
Allow the encoder to select between 'above', 'left', or combined 'above' and 'left' templates for template-based coding tools, signaling the chosen template type in the encoded data for the decoder to use the appropriate template for prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional L-shaped template is used for template-based coding tools, then the template structure is simple and easy to implement, but it fails to effectively capture the statistical properties of blocks when distorted by edges in non-camera captured content
Solution Approach 1:
The patent divides the conventional L-shaped template into two separate templates: a horizontal template for capturing statistics from adjacent pixels above the current block, and a vertical template for capturing statistics from adjacent pixels at the left of the current block. This segmentation allows each template to focus on capturing statistics in its respective direction without being distorted by edges, thereby improving the accuracy of block statistic capture while maintaining manageable complexity through modular structure.
2Adaptability or versatility
If multiple template types are introduced to improve coding performance, then the adaptability to different video content scenarios is improved, but the signaling overhead and decoder complexity increase
Solution Approach 1:
The patent introduces a template type indicator that dynamically selects between different template types (horizontal, vertical, or combined L-shaped) based on the characteristics of the current block. This dynamic adaptation allows the encoding system to choose the most appropriate template structure for each block, improving versatility for different video content scenarios. The complexity is managed by using a compact indicator mechanism that provides flexibility without excessive signaling overhead.
Data Source
AI summary
A video coding system may use different types of templates when using template-based prediction modes such as intra template matching prediction mode. The ‘L-shape’ template conventionally used by template-based coding tools may be split into a horizontal ‘above’ (or ‘upper’) template and a vertical ‘left’ template. This allows the encoder to select the appropriate part of the template that better captures statistical properties of the current block and results into improving the coding performance. The use of the type of template is signaled in the encoded data so that the decoder can use the appropriate template for the reconstruction when using a template-based prediction mode such as intra template matching prediction mode.