Research Of Multi-scale Edge Detection Based On Mathematical Morphology

In image processing, edge detection is a low-level visual processing, but due to the edge of the shape description of the basic characteristics of the target, making the edge detection in target recognition and image understanding to solve high-level image processing problem has a very important position. At the same time, edge detection is the classic image processing techniques in one of the problems, because the image from the noise in the imaging process, projection, mixing occurs as a result of image blur and distortion, making the edge is often difficult to detect. Therefore, the main surface of edge detectionConfronted with two questions: First, how to reduce noise interference; the second is how to maximize the effective edge detection information.The traditional method is to do image smoothing to reduce noise interference, has yielded some results, but not ideal, and in the filter out the noise, but also lost some edge information. In addition, the traditional single scale methods can not fully obtain the target image edge information.In recent years, with the scale space theory of depth, and some modern mathematical methods in the extensive application of computer vision, more in line with characteristics of human visual detection methods and multi-scale mathematical morphology edge detection, edge detection by scholars in the field their attention. View of this, to scale space theory and the mathematical morphology theory, mathematical morphology-based multi-scale edge detection method of inquiry.The paper focuses on the basic mathematical morphology edge detection, edge detection-type anti-noise and multi-structured edge detection method, and found by experiment: If you use multiple smaller geometry morphological structuring elements, morphological structure of the edge multi- detection method can detect the edge details of goals, but poor noise immunity. Traditional method of drawing a smooth filtering process by adding the idea of the structure of the above method is improved more than that in the expression of a geometric structure by adding a larger filter structure elements, the experiment has achieved good detection.In addition, the paper also use of filtering structure element of the self-expanding scale for morphological classification, the method is extended up to scale way to make more visual features. The experiments show that the improved multi-scale method can not only reduce noise, and better maintain the edge. And these details show that the method is feasible and effective.