Some Improvement And Application Researches On Edge Detection Methods Of Digital Image

As one of the most important feature of the digital image, edge's precise detection has very profound meanings of guidance in people's understanding and applying images. On condition of having summarized and analyzed the existing edge detection methods, this paper combined some development tools such as visual c++ and matlab, discussed basical edge detection method and several processes of its improvements from the point of enhancing the detection performance, and then proposed a synthetical method with nice property, thus provided new thoughts and approaches for raising new methods and strengthen the existing ones. Finally researched on the applications of edge detection in several various occasions, and put forward satisfactory solutions for each one of them.Chapter one stressly reviewed the edge detection methods we had on the base of analyzing relative background of image processing and edge detection. Indicated several problem questions that the technology had, expanding to the main research questions and arrangements for the content of this thesis.Chapter two effectively and rationally divided the general procedures of edge detection. On the base of that, deeply discussed, realized and analyzed kinds of image enhancement such as gray mapping, spatial domain filtering, frequency domain filtering and detection operator, point out the advantages, disadvantages and applicative objects of each treatment, which laid the foundation for the following improvement studies.Chapter three launched research on some aspects contained antinoise capability, location accuracy, detection detail and edge morphology. On antinoise capability, raised some modifications of median filtering and omnibearing one of morphological filtering, and did some research with optimal filtering based on canny's formula; On location accuracy, first analyzed that of edge detection operators, in turn discussed accurate positioning based on zero crossing thought, and led to the concept of subpixel location after ulteriorly advanced the location accuracy of LoG operator; On detection detail, respectively put contraposing solutions forward towards different application requirements, for instance, specific edge detail detecting, multi-scale edge detail detecting and secondary outlying edge detail simplifying; On edge morphology, carried on improvement research on three edge's aspects like continuity, fineness and smoothness, and what's more, proposed the way of combining threshold segmentation and edge detection operators, which obviously ameliorated the detected edge's continuity and fineness.Chapter four proposed a high-performance rapid controllable edge detection method lying on the research of previous two chapters. The method integrated laplacian template and region growing thought, adopted a modified operator template to heighten the border, controlled the seed edge pixels' growing to obtain the edge, and supported it with optimal disposes in every link, sequentially attained the performance which could be compared with canny, and had advantages over the latter. Raise an evaluation protocol of detecting effectiveness towards the edge's respondence and continuity, so lots of methods could be contrasted quantitatively, and the proposed one could be proved.Chapter five deliberated the specific applications of edge detection in engineering practice. Designated cellular edge's morphology detection, vehicle licenses automatic identification and moving object's tracking and detection from numerous application domains, and gave out kinds of targeted integrated approaches to resolve them effectively, which acquired satisfactory results. Chapter six concluded the researches of this issue, indicated certain innovations, and prospected numbers of prospective matters and directions need to be studied.