| Edge detection plays an important role in the digital image processing. Theconventional edge detection operators, such as Roberts, Sobel, Prewitt, Laplace,which are linear filtering methods, and miss somewhat in the results of edge detection,furthermore they are sensitive to noise. Mathematical morphology, a methodology ofnonlinear image processing and analysis theory, it has a complete set of theories,methods and algorithm system. It has some morphological characteristics such assimplifying image data, maintaining the basic shape of the image morphologicalcharacteristics, the study object of mathematical morphology is morphologicalcharacteristics of image. So it has widely used in various fields of image processing.This artical firstly explores and studies the ideas and nature of conventional edgedetection operators(ie: first-order operators such as Roberts, Sobel, Prewitt,second-order operators such as Laplace, Kirsch, Robinson and Log) and the new edgedetection operators(ie: Canny, wavelet edge detection algorithms and edge detectionalgorithm based on fuzzy theory)，simultaneously gives the experimental results ofthese operators to analyze the advantages and disadvantages. Deeply discusses thebasic theory of mathematical morphology from the the binary morphologicalpromotion to grayscale morphology, analysis and introduce the Mathematicalmorphology in the image processing applications.Secondly, based on the conventional morphological edge detection operators, thisartical improves and amends the existing anti-noise operators, combined withmulti-structure elements and multi-scale concept, put forward a new anti-noisemorphological edge detection algorithm of multi-structure and multi-scale elements.Finally, for the problem of pavement crack image edge detection, this artical putsforwards morphological edge detection algorithm combined with the improvedmedian filtering, meanwhile experiments show that the algorithm is more feasible andhas good application value. |
PDF Full Text Request