Leather Image Segmentation And Processing

The leather image is a special kind of image, and has some obvious characteristics different from the others. For example: 1) it is full-texture image; 2) when arranging the templates and cutting materials, as we cannot cut leather so smoothly, it has many anomalous "thorn" on its boundary, whose area is too small to be used; 3) in practice, people can only use the part inside of the leather, so we can only exceed the edge of the leather little or even cannot; 4) we should save the leather as possible as we can in the course of arrangement.Contraposed those characteristics of leather, after studying all kinds of algorithms of image segmentation, this dissertation chooses the techniques of thresholding methods of parallel region segmentation and edge tracing methods of serial edge segmentation to detect the contour of the leather image. Then by using polygonal approximation method, we can convert the contour's form from image into graphics. It will make following arrangement works more convenient. In this dissertation, we also use morphological opening operation to get rid of particle noise and reduce the effects of the tiny "thorn".The new ideas of this dissertation mainly lie in two points. Based on the research on the traditional edge tracing techniques, this dissertation presents an improved method according to LML rule. At the same time, studying the polygonal approximation methods, this dissertation also presents an improved and based on mergence method to image contour, and implements the transform of the contour from image to graphics preferably.