Study Of The Thinning Algorithms Of Line Drawing Image

Thinning can be defined as the successive erosion of the outermost layers of a figure until only a connected, unit width framework or 'skeleton' remains. This skeleton runs, ideally, along the medial lines of the limbs of the figure. Since the idea was first developed twenty years ago, thinning has become an important process in pattern recognition. In surveying and mapping field, the binary image's thinning is an improtant content of the map vector, and also is the base of the map's symbol recognition. In edge tracking and classification of the remote sensing image, the thinning of the threadiness images are important constituent parts.The biggest difficulty in devising a successful thinning algorithm is taking account of global properties of the patterns. Added to this is the difficulty of determining whether a point is an incipient 'noise' and whether, therefore, it should be retained or eliminated. According to different purposes of the applications we must choose different thinning methodes, these are due to the types of the thinning objects. Thus, the studying on thinning methodes based on practical applications has particular realistic significance.On the bases of discussing on the fundamental principles of thinning, and the preprocessing prior to thinning, an analysis was made on the factors in the process of thinning the line drawings. And after that, a detail analysis was made on two of the factors of thinning (acute angle and singular pixel). For the error of the acute angle's thinning, an analysis was made about the error factors of thinning, proposing the improved program in the preprocessing, and then analysis the problems of stroke trend and the "Lump" in the thinning methods. For the problem of the singular pixel, proceed with the operator of pixel deletion after the thinning results to guarantee the singular pixel of the skeleton.There's discussing about some primary thinning methodes in this paper, then making some improvements in the program of Zhang-Suen. There are some comparisons between the original program and the processed program at last.