Premium
Rotation invariant image template matching based on correlation of curvature distribution
Author(s) -
Saitoh Fumihiko
Publication year - 2003
Publication title -
electrical engineering in japan
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.136
H-Index - 28
eISSN - 1520-6416
pISSN - 0424-7760
DOI - 10.1002/eej.10235
Subject(s) - pixel , curvature , computer vision , artificial intelligence , mathematics , invariant (physics) , template matching , image (mathematics) , rotation (mathematics) , enhanced data rates for gsm evolution , matching (statistics) , computer science , geometry , statistics , mathematical physics
This paper proposes a method of searching for a target image area that has an arbitrary location and inclination in an objective image. The method uses the curvature of the edge pixels, which is invariant to the location and inclination of a target image area. The curvature of an edge pixel and the distance between the central pixel and the edge pixel are measured in a circular image area. The stored value in the two‐dimensional space that is spread by the curvature and the distance is incremented. This process is repeated about all edge pixels in the circular image area. The matching is executed between the two‐dimensional spaces that are generated by the circular template image and generated by a circular image area in an objective image. The experimental results show that the recognition rate of the target image areas recorded 97.8% of four kinds of images. The averaged processing time to search the target image areas was 20.4 seconds. © 2003 Wiley Periodicals, Inc. Electr Eng Jpn, 145(4): 56–3, 2003; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.10235