Rotation Invariant Image Matching Based on Correlation of Variance of Local Edge Directions

Abstract

Image template matching has been applied to various applications because it can search a similar partial image area to a registered template image in an objective gray-scaled image. This paper proposes a method to search a target image area that has free location and free inclination to a template image. The proposed method uses the variance of local edge directions that is an invariant parameter to the changes of location and inclination of an objective image pattern. High speed matching is expected because the method uses only local information around each edge pixel. The experimental results show that the averaged recognition rate about four kinds of experimental images recorded 99.6%. The proper target image areas were searched in the objective images with free locations and free inclinations. The averaged processing time was 0.16 second.