2007 年 73 巻 9 号 p. 1051-1056
In order to realize highly sensitive image comparison, a method for robust image matching, which is not affected by the form or the density of the patterns contained in the images, is required. Thus, a method that utilizes the characteristics of a normalized correlation coefficient is proposed in this study. Using this method, an object image obtained under active illumination control is subdivided into multiple smaller images, and the correlation coefficient maps are individually computed. The matching position is subsequently determined by integrating the maps. The effectiveness of the method is confirmed through experiments.