Translation, Rotation, and Scale-Invariant Template Matching

Abstract

new template matching algorithm that enables fast matching between a reference template and a scene image regardless of its translation, rotation and scaling is proposed in this paper. In the beginning, the ring projection transform process is used to convert the 2D template into a 1D signal. Then, the template matching is performed by constructing a parametric template vector of the 1D signal with differently scaled templates of the object. Our approach is conceptually simple, easy for implementation and very efficient for template matching (no iterative steps are involved). Experimental results have shown that the proposed algorithm is not only efficient and robust to detect the target object with the changes of translation, orientation and scale, but also can estimate the scaling value of the target object in the input scene.