Adaptive Motion Vector Search Range for Real-Time Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion vector estimation methods, such as FBMA and median prediction, are computationally complex and inefficient, especially in high-definition video encoding, leading to increased hardware requirements and reduced reliability due to limited use of neighboring macro block correlations.
Innovation Solution
Adaptive motion vector search range estimation based on the maximum and minimum magnitudes of neighboring macro blocks, reducing computational burden and improving reliability by dynamically defining the search range using a combination of neighboring block data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full block matching algorithm (FBMA) is used to search all macro blocks in a wide search range, then motion vector accuracy is improved, but computational complexity and processing time increase enormously
Solution Approach 1:
The patent segments the search process into two stages: first determining a limited search range based on neighboring macro block motion vectors, then performing exhaustive search only within this constrained range. This segmentation reduces the total number of comparisons from the entire frame to a localized region, maintaining accuracy while reducing complexity.
Solution Approach 2:
The patent performs preliminary action by estimating the motion vector of the current macro block using motion vectors from neighboring macro blocks before conducting the exhaustive search. This preliminary estimation defines a constrained search range, allowing the system to focus computational resources only on the most relevant area and reduce overall processing time.
2Reliability
If motion vectors of multiple neighboring macro blocks are used for prediction, then reliability of motion vector estimation is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by selectively using motion vectors from neighboring macro blocks (typically left, upper, and upper-right neighbors) rather than all available blocks. This selective approach provides sufficient reliability for estimation while avoiding the excessive computational burden of analyzing every possible neighboring block.
Solution Approach 2:
The patent applies local quality by focusing computational resources on locally relevant neighboring macro blocks that are spatially adjacent to the current block. This local approach assumes that motion characteristics are more similar between adjacent blocks, providing reliable estimation with reduced computation compared to global analysis.
3Measurement precision
If a wide search range is defined for high definition video, then motion vector accuracy is improved, but real-time processing becomes impossible due to enormous number of pixels to compare
Solution Approach 1:
The patent performs preliminary estimation of the motion vector using neighboring macro block data before conducting exhaustive search. This preliminary action defines a constrained search range that is sufficient for HD video quality while reducing the pixel comparison workload to manageable levels, enabling real-time processing.
Solution Approach 2:
The patent applies dynamics by adaptively determining the search range based on the actual motion characteristics of neighboring blocks rather than using a fixed large range. This dynamic adjustment allows the search range to expand or contract based on scene content, maintaining accuracy for high motion scenarios while reducing computations for low motion scenarios, thus enabling real-time processing.
Data Source
AI summary
The video encoder in accordance with the present invention divides input image signal into macro blocks, estimates motion vectors of each macro block and finally encodes the input image signal. The encoder determines the number of macro blocks adjacent to a current macro block. If the number of adjacent macro blocks is equal to or more than two, the encoder calculates a motion vector of the adjacent macro blocks and selects a macro block that has the largest motion vector. Then it defines a least search area that the current adaptive motion vector can have, and compares the least search area with the motion vector of the largest adjacent macro block, and finally determines the largest value as the search area of the adaptive motion vector. Next, the encoder compares the search area of the adaptive motion vector with the search area of the user-defined motion adaptive vector to thereby determine the least value as the search area of the final adaptive motion vector.