Motion Vector Reliability via Similarity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for calculating motion vectors often result in low precision and accuracy, particularly in monotonous image areas with few features, and can include incorrect vectors, leading to inefficient estimation processes.
Innovation Solution
An image processing apparatus that acquires motion vectors from consecutive images, selects neighboring vectors, calculates their similarity, and determines reliability based on the number of neighboring vectors with high similarity, excluding low-precision vectors to reduce outliers and iterations in robust estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion vectors are calculated using block matching or feature point matching in monotonous image areas, then the degree of similarity is calculated as high, but the precision and accuracy of motion vectors deteriorate
Solution Approach 1:
The patent introduces an intermediary verification step by comparing motion vectors from multiple different feature points within the same block. This intermediary comparison acts as a mediator to detect and eliminate incorrect motion vectors that may arise in monotonous areas, thereby resolving the contradiction between high similarity scores and actual vector accuracy
Solution Approach 2:
The patent calculates motion vectors from multiple different feature points (creating copies of the measurement process) within the same block. By comparing these copied vectors, the system can identify outliers and incorrect vectors, thus improving measurement precision while maintaining the simplicity of block matching methodology
2Measurement precision
If robust estimation is performed with incorrect motion vectors included, then the estimation process becomes more complex, but the precision of motion parameters deteriorates
Solution Approach 1:
The patent performs preliminary filtering of motion vectors by comparing vectors from multiple feature points before conducting robust estimation. This preliminary action removes incorrect vectors in advance, reducing the number of outliers and thereby simplifying the subsequent robust estimation process while improving the precision of motion parameters
Solution Approach 2:
The patent extracts and removes incorrect motion vectors from the dataset before performing robust estimation. By taking out the harmful incorrect vectors through comparison and identification, the system reduces the complexity of the estimation process and improves the precision of the resulting motion parameters
3Device complexity
If feature points are concentrated in specific areas of images, then the calculation is simplified, but the reliability of motion vectors deteriorates
Solution Approach 1:
The patent divides the image into multiple blocks and further segments each block into multiple feature points. This segmentation ensures that motion vectors are calculated from multiple distributed locations within each block, improving the reliability of motion vectors while maintaining computational simplicity through the structured block-based approach
Data Source
AI summary
A plurality of motion vectors are acquired from consecutive images. From the acquired plurality of motion vectors, a motion vector of interest and neighboring motion vectors neighboring the motion vector of interest are selected and a degree of similarity between two motion vectors is acquired. A value related to the total number of neighboring motion vectors having high degrees of similarity to the motion vector of interest is acquired as a degree of reliability.