| Edge detection based on mathematical morphology is a great important researcharea in image analysis. Typically, set theory is the core of morphology and we usuallyuse eroding, dilating, opening and closing computation to complete image processing.According to processing methods of set theory, the efficiency of image edge detectioncan be improved greatly. This paper combining morphology with edge detectionalgorithm outweighs some algorithms, such as Roberts、Sobel、Prewitt、LoG、Canny.Firstly, the common computations of mathematical morphology are analyzed inthis paper. Using these computations effectively, satisfying pre-processing image、district filling and edge thinning can be acquired.Secondly, noise filtering in digital image processing is an essential operation.Usually, during the process of transferring and storing of image exterior noise can bebrought in original image. In order to solving this problem, many algorithms alwaysadopt Gauss function to reduce noise interference as Canny algorithm. Therefore, facingto high level noise (60%), traditional edge detection methods cannot produce excitingedge information, so pre-processing is important. The third chapter will research noisefiltering algorithms specially and the novel filtering method is given by the author.For researching extensively, this paper compares the disadvantages and advantagesof different edge detection methods. Robers、Sobel and Prewitt and other algorithmsbased on them are all one dimensional methods and normally noise sensitive, whileLoG and Canny methods are all two dimensional methods with an improved noisefiltering by the Gauss filtering.At last, this paper proposes a novel edge detection method with the knowledge ofmathematical morphology and improved Canny. It is proposed by three flaws oftraditional Canny algorithm. Before edge detection, we use the noise filtering methodproposed by author to acquire a satisfying input image and mathematical morphology tomake edge thinning and restoring image. |
PDF Full Text Request