Adaptive Hierarchical Phase Correlation for Motion Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion estimation methods face challenges in achieving accurate results due to fixed threshold values in phase correlation-based methods, leading to suboptimal Peak Signal to Noise Ratio (PSNR) values and increased computational load, particularly when dealing with varying motion components and unequal frame sizes.
Innovation Solution
The method employs adaptive threshold calculation and hierarchical phase correlation with iterative splitting and overlapping of blocks to determine the partitioning of frames, ensuring reliable data collection and reducing the need for interpolation, thereby enhancing PSNR and reducing computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed threshold value is used in partition criteria, then the algorithm is simple to implement, but accurate results cannot be obtained in some cases due to inability to adapt to varying motion components
Solution Approach 1:
The patent applies dynamics by making the threshold value adaptive rather than fixed. The threshold automatically adjusts based on the characteristics of the phase correlation surface and motion components in each specific case, allowing the algorithm to adapt to varying motion scenarios and achieve accurate results without requiring complex manual tuning.
Solution Approach 2:
The patent changes the parameter threshold from a fixed constant to a dynamic value that varies according to the input data characteristics. This parameter change enables the algorithm to optimize its performance for different motion patterns while maintaining computational efficiency.
2Measurement precision
If iterative partitioning is applied to achieve accurate motion vectors, then measurement precision improves, but computational load increases
Solution Approach 1:
The patent applies segmentation by dividing the frame into blocks and further partitioning them into quadrants based on motion characteristics. This hierarchical segmentation allows the algorithm to focus computational resources on regions with complex motion while using coarser resolution for uniform regions, achieving accurate motion vectors with reduced overall computational load.
Solution Approach 2:
The patent applies local quality by applying different levels of partitioning and processing to different regions of the frame based on their local motion characteristics. Regions with single dominant motion use coarser partitioning while regions with multiple motion components undergo finer partitioning, optimizing the balance between accuracy and computational efficiency.
3Adaptability or versatility
If unequal size image blocks are correlated, then the method can handle different frame sizes, but interpolation is required which increases computational requirements
Solution Approach 1:
The patent applies asymmetry by allowing reference and target blocks to have different sizes without requiring symmetric interpolation operations. The phase correlation method inherently handles asymmetric block sizes by working directly with the Fourier transforms of blocks of different dimensions, eliminating the need for complex interpolation to equalize sizes.
4Reliability
If overlapping area is extended in hierarchical partitioning, then reliable data is obtained for phase correlation, but the processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the Fourier transforms of blocks at different hierarchical levels before performing phase correlation. This preliminary computation of overlapping regions allows the algorithm to efficiently access pre-computed data during motion estimation, improving reliability while reducing the complexity of repeated calculations.
Data Source
Figure 1~2
Figure 3~4(b)
Figure 5~6
AI summary
The motion data in a video is a very important item to save or transmit the video. It decreases the storage area and transmission bandwidth. So the motion estimation issue becomes very important in the video technology. The present invention about motion estimation is based on the frequency domain and it focuses on the phase correlation of the frame sequences. The main idea behind this invention is to compare the ratio between the first highest peak and second highest peak with an adaptive threshold value which changes itself according to the result of the phase correlation. If the ratio is lower, there are two or more motions included in the block and it should be further split. Then the splitting procedure continues until the size of blocks becomes equal to appropriate size. The motion vector corresponding to the highest peak is assigned to the entire frame. The new method provides a better PSNR performance with less motion vectors and automatically adapts itself to the frames.