抄録
A novel method for image template matching with the robustness for various real-world disturbances is proposed. The method is based on statistical reach feature method. The method utilizes the set of signs of differences for selected pairs of points from the template image. The selection is performed from the statistical view-point. The concrete description of the method is presented plainly. The selection of thresholds in the method is argued. Experiments with various disturbances including illumination change, occlusion, and JPEG noise are shown. The analyses based on the noise models are performed to show theoretical background of the robustness of the method.
