This paper presents a new approach including passive and actriveprocesses to deal with the image segmentation and pattern recognition to a color blindness plate (CBP). The CBP is well-known satisfactory way of testing the degree of color blindness happened in the human visual system. The image of CBP is very complex. It includes not only the colors but also the disconnected size-varied dots. It is very difficult by using a conventional machine vision algorithm to recognize the meaningful pattern (e.g., a figure) from such a CBP image. The proposed machine vision algorithms provide a new solution to this problem.