The Research And Application Of The Image Boundary Extraction Based On Pixel Coverage Segmentation

With the development of digital image processing, image edge detection technology is more and more appreciated by people, and has become one of the most important contents of digital image processing. In terms of visual, people often tell the difference between various objects by the edges, while in the image field, the edge is the most important feature what separates a region with another. As a result, image edge detection is widely used in image segmentation, identification, and other directions. Many scholars at home and abroad have been doing research in this area, but edge detection still faces many problems and challenges. The research of this paper is based on pixel coverage segmentation, which is around soft boundary and location information, and the main work includes:1. Image Coverage Segmentation Based on Soft Boundaries. This paper studies image object coverage segmentation by introducing soft boundaries. By using soft boundaries, fuzzy image can be segmented into several classes with a sharing boundary which is called a soft boundary. In this paper, several concepts of boundaries are defined, namely, hard boundary, inner boundary and outer boundary. Soft boundary is defined by the subtraction between inner boundary and outer boundary of a set. Coverage segmentation algorithm and optimization method are proposed in this paper. Meanwhile, neighbor decision rules are used in classification of pixels to filter noise or outliers. Experiments and comparison with classical coverage segmentation methods are presented, including noise test on the proposed method with four kinds of boundaries and neighbor decision rules.2. Coverage Segmentation Based on Location Information. This paper studies a coverage segmentation method based on location information, which has a good performance on fuzzy boundary images, comparing to the previous method which only focus on crisp objects. It is start by pre-segmenting the fuzzy boundary image, used Multi-directions Algorithm to detect pixels intersected with the boundary of continuous imaging objects, and re-distribute coverage values for the possible mixed pixel set based on an improved coverage segmentation method. Experiments and comparison with classical coverage segmentation methods are presented, including noise test on the proposed method with four kinds of boundaries and neighbor decision rules. The results based on a publicly available images suggest that our method potentially can be more accurate than comparable state-of-the-art methods proposed in literature.3. Image Boundary Extraction Based on Pixel Coverage Segmentation and Chan-Vese Model. This paper studies a coarse-to-fine approach for image boundary extraction, which is mainly rely on pixel coverage segmentation and Chan-Vese model, which has a good performance on fuzzy boundary images, comparing to the previous method which only focus on crisp objects. In this paper, coverage segmentation algorithm and refinement based on activecontours method are proposed, which is start by the pixel coverage segmentation. And then, the image border post-detection is refined using the active-contours algorithm. Experiments and comparison with classical methods are presented. The results based on a publicly available images suggest that our method potentially can be more accurate than comparable state-of-the-art methods proposed in literature.