1999 年 65 巻 6 号 p. 820-824
A new method to precisely detect coordinates of the edge position in a blurred image is proposed. Stereo vision, techniques are widely used to measure the dimension of objects. However, it is difficult to apply the stereo vision techniques to small objects, because the most area of images taken through microscopes are blurred due to the narrow depth of field and the accuracy of the edge detection in those images is deteriorated. In order to obtain the position of the blurred edges with high accuracy, ray tracing numerical simulation method of blurred images is developed and an inverse analysis using the simulation is employed to detect the edge position. Some experiments using mono-vision probe and stereo-vision probe are performed using a simple specimen. As a result, maximum error of x-z coordinates by stereo vision is less than the size of one CCD pixel (3.8μm) over the any part of images. It is found that precise coordinates detection of edge position is possible through inverse analysis even though the images are blurred.