Fast Image Matching via Angle Prediction Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image matching techniques, such as brute force methods, are time-consuming and inefficient due to the need for extensive data comparison, making them unsuitable for fast image matching applications.
Innovation Solution
The method involves creating an angle prediction table by pre-comparing template images with various angular orientations, allowing for the rapid identification of possible tilt angles in testing images by looking up the table, thereby reducing the number of necessary comparisons and enhancing matching speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If brute force method is used to compare template images with testing images at all angular orientations, then image matching completeness is improved, but processing time increases significantly
Solution Approach 1:
The patent pre-computes and stores angle prediction tables that map similarity values to corresponding angular orientations before actual image matching occurs. During processing, these pre-generated tables are queried to quickly determine candidate angles, eliminating the need to evaluate all 360 degrees systematically. This preliminary preparation significantly reduces real-time processing time while maintaining matching reliability.
Solution Approach 2:
The patent introduces angle prediction tables as intermediary data structures that mediate between the template image and testing image comparison. These tables serve as a lookup mechanism that translates similarity measurements into candidate angular orientations, acting as an intermediary that bridges the gap between complete evaluation and fast processing requirements.
2Measurement precision
If extensive data comparison is performed to ensure accurate matching, then matching accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts and pre-computes the angular orientation information from the template image comparisons and stores it in angle prediction tables. This extraction separates the computational heavy lifting (pre-computation) from the actual matching process (lookup and verification), reducing the complexity of real-time processing while maintaining accuracy through the pre-verified angle data.
Solution Approach 2:
By performing the complex data comparison and angle determination in advance during table generation, the patent shifts computational complexity from runtime to setup time. The actual matching operation then becomes a simple table lookup and verification process, significantly reducing operational computational complexity while preserving matching accuracy.
Data Source
AI summary
Disclosed herein is a method of fast image matching that includes the steps as follows. A template image with a predetermined angular orientation is compared with template images in the range from 0 to 360 degrees to create an angle prediction table. Next, a testing image is acquired and compared with the template image with the predetermined angular orientation to record the similarity at each position, and a plurality of angles corresponding to the similarity is found from the angle prediction table. Afterwards, the template images of the plurality of angles are respectively compared with the testing image to obtain the highest similarity as a comparison result of the position.