Boundary Tracing,Region Filling And Applications Of Chain Code

Boundary tracing and region filling are basic problems in digital image processing. Transforming chain code from one to another is convenient by transformation algorithms. It's the important method to obtain geometric characteristic of image by chain code. The slant correction and form recognition of document image are the most important applications in the field of Optical Character Recognition.Research in this paper includes boundary tracing, transformation algorithms among chain code, region filling algorithm, obtaining geometric characteristic of image and form processing software based on chain code, algorithm and application of chain code are researched widely. The achievements are listed below:1. Algorithms are proposed for tracing boundary and extracting vertex chain code and obtaining the tree structure of contours from 8-adjacent and 4-adjacent images. There are two algorithms in all. Firstly, trace region boundary and obtains chain code to construct tree structure of contours through a pixel vertex matrix. Secondly, designed automatic tracing machine to obtain vertex chain code and tree structure of contours at the same time of tracing the boundary of the image. Both the algorithms are linear and can be applied to arbitrary complicated image region.2. This paper researches several transformation algorithms among different chain codes of binary image on square lattice. These algorithms are as follows: transformation algorithms of Freeman crack code and VCC, transformation algorithms of 4-direction and VCC, transformation algorithms of 8-direction and VCC. Once one kind of chain code of an image is obtained, the image information can be shared by other chain codes.3. Region filling algorithms based on chain code are analyzed and improved, including the region filling algorithms based on Freeman chain code, crack code, vertex chain code and a new parity matching algorithm. Finally, complexity analysis of each proposed algorithms are given. Experiments indicate that compared with the existing ones, these new algorithms work better. They work much better than the current methods especially in the case of multi-connected region or the whole image, because the proposed algorithms don't fill the holes in the region.4. In the paper, a coordinate labeling automata and an algorithm computing geometric moment and Euler number of image using vertex chain code are proposed.