Research Of Edge Detection Algorithm Based On Wavelet Transform And Morphology

In image processing, the edge detection is a low-level visual processing. But due to the edgeof the shape description of the basic characteristics of the target, the edge detection has a veryimportant position in high-level image processing problem, such as target recognition and imageunderstanding. And the edge detection is the classic problem in image processing techniques,because in the imaging process, the noise, projection and mixing make image blur and distort,which makes the edge detect difficultly. In recent years, with the development of the scale spacetheory, and the extensive application of some modern mathematical methods in computer vision,the multi-scale detection methods in line with characteristics of human visual and mathematicalmorphology edge detection are paid attention by scholars in edge detection. This paperresearches the application of mathematical morphology and wavelet multi-scale method in edgedetection.Firstly, this paper introduces the definition of image edge detection and the traditional edgedetection operators, analyzes the advantages and disadvantages of various operators, andvalidates it by experiment. Next, the basic theory, the development process and the currentresearch status of the mathematical morphology are introduced, the binary morphology and graybasic morphological operator: erosion, expansion, open operations and closed operations aregiven, and the all-round morphology filter is improved. The edge detection algorithm based onall-round morphology filter and circular all-round morphology filter is researched, and validatedby experiment. Finally, combining the wavelet transform with mathematical morphology, amorphological wavelet edge detection algorithm is given. In this algorithm, the image isdenoised by wavelet soft threshold method, then the edge is detected by multi-functionalstructure element template, and the different wavelet de-noising experiment results are compared.The experimental results show that the edges detected have good definition. Finally, the wholedissertation is summarized and expected.