Decoder Side Motion Vector Derivation Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding and decoding technologies face challenges in efficiently reducing redundancy in video signals, particularly in motion vector prediction, which can lead to increased bandwidth and storage requirements due to limitations in existing motion vector derivation and refinement methods.
Innovation Solution
The implementation of decoder-side motion vector derivation and refinement using template matching, where a template of a first block is determined and used to derive motion information from a second block, allowing for more efficient motion vector prediction and compression by leveraging reconstruction samples from previously decoded regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vector prediction is performed using conventional methods, then bandwidth requirements are reduced, but prediction accuracy is insufficient leading to higher distortion
Solution Approach 1:
The template for motion vector prediction is segmented into multiple regions, with different weighting factors applied to different regions. This allows the decoder to use a more accurate template matching approach by considering multiple spatial regions with appropriate weights, improving prediction accuracy without requiring additional bandwidth for transmitting enhanced motion information
Solution Approach 2:
The invention changes the parameters of template matching by introducing region-specific weighting factors and using reconstruction samples from previously decoded blocks. This modifies the traditional template matching process to achieve higher prediction accuracy while maintaining the same bandwidth constraints, as the improved accuracy comes from smarter parameter usage rather than more data transmission
2Measurement precision
If template matching uses reconstruction samples from previously decoded blocks, then prediction accuracy improves, but device complexity increases
Solution Approach 1:
The invention performs preliminary actions by using reconstruction samples from previously decoded blocks that are already available in the decoder buffer. This eliminates the need for additional complex processing to generate reference samples, as the preliminary decoding of previous blocks provides the necessary data. The template matching process leverages this pre-computed information to improve accuracy without adding significant complexity
Solution Approach 2:
The decoder serves itself by using its own previously decoded reconstruction samples as the template for motion vector prediction. This self-service approach eliminates the need for additional reference data from external sources or complex inter-processor communication, reducing device complexity while improving prediction accuracy through effective utilization of already-available data
3Loss of information
If more motion vector data is transmitted to improve prediction accuracy, then distortion is reduced, but bandwidth requirements increase
Solution Approach 1:
The invention introduces an intermediary template matching process that uses reconstruction samples from previously decoded blocks as an intermediate step between the encoded data and the final decoded output. This intermediary process improves prediction accuracy by providing better motion compensation without requiring additional motion vector data to be transmitted, as the template matching derives improved motion information from the available reconstruction samples
Solution Approach 2:
The invention creates a copy of the template from previously decoded reconstruction samples and uses this copied template for matching against the current block. This copying approach allows the decoder to leverage already-decoded data for improved prediction without requiring additional transmission of motion information, reducing bitrate requirements while maintaining prediction accuracy
Data Source
AI summary
A method of video decoding in a decoder is described. A coded video bit stream is received. A template of a first block in a picture in the received coded bitstream is determined in response to a determination that decoder-side motion vector derivation (DMVD) is applied to the first block for reconstruction. A first region of the template is determined to be overlapping with a second block that is decoded prior to the first block. Further, a second region of the template is determined by subtracting the first region from the template. Motion information of the first block is derived using template matching based on reconstruction samples of the second region.