Template Matching Prediction with Multiple Template Types for Screen Content
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in efficiently compressing and transmitting large video sequences due to the high data size, which can be exacerbated by the complexity of screen content and synthetic graphics, leading to inefficiencies in storage and transmission.
Innovation Solution
The implementation of template matching prediction (TMP) modes, including intra and inter prediction techniques, to enhance video encoding and decoding processes, particularly for screen content, by utilizing multiple template types and adaptive block partitioning strategies, such as quadtree+multi-type tree partitioning, to improve compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching prediction with multiple template types is implemented, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments the prediction process by dividing the current block into multiple sub-blocks and applying different template matching strategies to each sub-block. This allows the system to achieve high compression efficiency through selective application of complex techniques only where needed, rather than uniformly across the entire block, thus managing device complexity while maintaining productivity.
Solution Approach 2:
The patent implements dynamic template selection where the encoder adapts the template matching approach based on local block characteristics. Different template types (e.g., rectangular, triangular, irregular shapes) are dynamically selected and applied to different regions based on their specific features, optimizing compression efficiency while avoiding unnecessary complexity in uniform regions.
2Manufacturing precision
If adaptive block partitioning strategies are used, then video quality is maintained, but loss of time increases
Solution Approach 1:
The patent applies local quality principles by using adaptive block partitioning that divides the video frame into regions of different sizes and complexities. High-quality template matching is applied to complex regions requiring detailed representation, while simpler regions use coarser partitioning. This maintains overall video quality while reducing the total computational time required compared to uniform fine-grained partitioning.
Solution Approach 2:
The patent employs partial action by selectively applying complex template matching and adaptive partitioning only to blocks that require it based on complexity metrics. Not all blocks undergo the full processing pipeline - some are handled with simplified methods, reducing total encoding time while maintaining quality where it matters most.
3Productivity
If multiple template types are utilized, then compression efficiency is improved, but loss of information increases
Solution Approach 1:
The patent uses preliminary action by pre-defining a set of common template types and their associated parameters. During encoding, the system selects from these pre-established templates rather than creating entirely new ones, which reduces the information overhead required to describe the templates. The preliminary classification of block types allows efficient selection with minimal additional bitrate.
Solution Approach 2:
The patent employs copying by reusing template patterns across multiple blocks when appropriate. Once a template type is defined and its parameters established, the same template can be copied and applied to similar blocks with minimal additional information needed, just the template index and transformation parameters rather than full re-description.
Data Source
AI summary
A decoder searches first candidate templates in a first search region and second candidate templates in a second search region to select a reference template based on calculating template matching (TM) costs comprising first TM costs of the first candidate templates and second TM costs of the second candidate templates. The first candidate templates, of first candidate reference blocks (RBs) from the first search region, each corresponds to a current template of a current block. The second candidate templates, of second candidate RBs from the second search region, each corresponds to the current template flipped in a direction. The current block is decoded based on a RB indicated by the reference template.