Video Decoder Motion Estimation Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding technologies, such as High-Efficiency Video Coding (HEVC), face challenges in reducing the processing load and memory bandwidth required for motion estimation and memory access during the decoding process, particularly due to the need to search for similar regions in large reference pictures.
Innovation Solution
A decoder and encoder system that determines a motion estimation region within a reference picture, allowing motion estimation to be performed only within this region, thereby reducing the processing load and memory bandwidth by excluding candidates with motion vectors outside the region and using pattern matching within the adjacent region to determine the motion vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion estimation is performed by searching for similar regions in the entire reference picture, then motion vector accuracy is improved, but processing load and memory bandwidth increase
Solution Approach 1:
The reference picture is divided into a motion estimation region and a non-motion estimation region. The motion estimation region is determined based on candidate vectors from neighboring blocks, allowing the decoder to perform motion estimation only within this segmented region rather than the entire reference picture, thus reducing processing load while maintaining accuracy.
Solution Approach 2:
Different regions of the reference picture are treated differently: the motion estimation region is processed with full motion estimation to ensure accuracy, while the non-motion estimation region is excluded from processing. This local differentiation optimizes the balance between accuracy and processing load.
2Measurement precision
If motion estimation is performed by searching for similar regions in the entire reference picture, then motion vector accuracy is improved, but memory bandwidth increases
Solution Approach 1:
The reference picture is segmented into a motion estimation region and a non-motion estimation region. By restricting motion estimation to only the motion estimation region, the amount of data that needs to be accessed from memory is reduced, thereby decreasing memory bandwidth requirements while maintaining motion vector accuracy.
Solution Approach 2:
The non-motion estimation region is extracted and excluded from the motion estimation process. Only the motion estimation region is processed, which reduces the quantity of data that needs to be loaded into memory during motion estimation, thus reducing memory bandwidth consumption.
3Measurement precision
If the search region for motion estimation is enlarged, then motion vector accuracy is improved, but processing time increases
Solution Approach 1:
The reference picture is segmented to identify a motion estimation region that is sufficient for accurate motion vector derivation. By using this segmented region rather than the entire reference picture, the processing time is reduced while maintaining the accuracy needed for proper motion compensation.
Solution Approach 2:
Instead of performing motion estimation across the entire reference picture (excessive action), the patent applies motion estimation only to the necessary motion estimation region (partial action). This partial approach achieves sufficient accuracy without the time cost of processing the entire picture.
Data Source
AI summary
A decoder that decodes a current block using a motion vector includes: a processor; and memory. Using the memory, the processor: derives a first candidate vector from one or more candidate vectors of one or more neighboring blocks that neighbor the current block; determines, in a first reference picture for the current block, a first adjacent region that includes a position indicated by the first candidate vector; calculates evaluation values of a plurality of candidate regions included in the first adjacent region; and determines a first motion vector of the current block, based on a first candidate region having a smallest evaluation value among the evaluation values. The first adjacent region is included in a first motion estimation region determined based on the position indicated by the first candidate vector.