Inter Prediction Candidate Lists Using FIFO Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in constructing candidate lists for inter prediction, particularly in advanced codecs like VVC, leading to increased computational complexity and reduced coding efficiency.
Innovation Solution
A method for constructing candidate lists using a first-in-first-out (FIFO) queue for historical inter prediction information, combined with template matching costs and scanner order, to determine the inter prediction information for current blocks, allowing for efficient selection of candidate blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional candidate list construction methods are used for inter prediction, then coding efficiency can be maintained with basic complexity, but computational complexity increases and coding efficiency decreases when using advanced techniques
Solution Approach 1:
The patent pre-calculates and stores template matching costs for candidate blocks in a FIFO queue during the encoding process. By preparing this information in advance, the decoder can efficiently select the best candidate without performing complex real-time calculations, thus improving coding efficiency while managing computational complexity
Solution Approach 2:
The system uses historical inter prediction information from previously decoded blocks to automatically construct candidate lists for current blocks. The FIFO queue mechanism allows the system to self-manage the candidate pool using its own decoded data, reducing the need for external complex processing while maintaining high coding efficiency
2Measurement precision
If more candidate blocks are included in the candidate list to improve prediction accuracy, then coding efficiency improves, but the complexity of managing and processing the candidate list increases
Solution Approach 1:
The patent changes the parameter of candidate block selection by using template matching costs as a filtering criterion. Instead of managing all possible candidate blocks, the system evaluates and ranks candidates based on their template matching costs, selecting only those that meet certain thresholds. This parameter-based filtering maintains prediction accuracy while significantly reducing the complexity of candidate list management
3Speed
If historical inter prediction information is stored and reused for multiple blocks, then processing speed improves, but memory requirements and data management complexity increase
Solution Approach 1:
The patent implements a FIFO (First-In-First-Out) queue mechanism for storing historical inter prediction information. When the queue reaches its capacity, the oldest entries are automatically discarded to make room for new candidates. This approach allows the system to maintain a manageable amount of historical data in memory while still benefiting from the speed improvements of reusing recent prediction information, effectively balancing memory requirements with processing speed
Data Source
AI summary
A method includes receiving a bitstream of coded information for a sequence of pictures, the coded information indicates an inter prediction of a current block in a current picture using a formular based inter prediction technique, the formular based inter prediction technique generates a prediction sample of the current block based on a formular with reconstructed samples of a reference block in a reference picture being input to the formular, the formular includes parameters derived based on a current template of the current block and a reference template of the reference block. The method also includes constructing a candidate list including one or more coded blocks associated with the current block, a coded block in the one or more coded blocks is a candidate that provides formular based inter prediction information.