Study On Edge Detection Method Of Images Based On Dyadic Wavelet Transform

Edge is one of the basic features of images, and it not only provides intuitive visual information for description, explanation and recognition, but also carries a lot of valuable information. The edge detection is the premise work and top priority of image analysis and understanding. The accuracy of test results directly affects image analysis and understanding of the computer vision system. Many methods have been presented in this field, however, because of the complexity of the edge detection and its inherent problem, they have limitations and deficiencies, especially in noise suppression and detection positioning. In recent years, dyadic wavelet transform as a new effective image processing tool has showed its unique advantages. Because dyadic wavelet transform doesn’t down sampling and has the translational invariance, it shows good performance in terms of the accuracy of the edge detection and can locate the image edge more accurately.Based on dyadic wavelet transform theory, this thesis studied the problem of the edge detection of images. Firstly, B-spline initial dyadic wavelet filters, the dyadic lifting schemes and two theorems which were proposed by T.Abdukirim are analyzed. Based on the dyadic lifting schemes, lifting dyadic wavelet filters with symmetry, compactly supported and higher number of vanishing moments are constructed starting from B-spline initial dyadic wavelet filters of r ?1, m ?4 and r ?2, m ?1by choosing appropriate free parameters. Secondly, by comparing and analyzing the advantages and disadvantages of typical method about edge detection, the author puts forward a new method of the edge detection of images which is based on dyadic wavelet transform theory, the original image is decomposed by àtrous to get horizontal low-frequency sub-image and vertical high-frequency sub-image at first. Then they are gradient modulus values by modified modulus formula. Finally, they are exploited by the morphological thinning technique. Results of lots of experiments show that this method can effectively balance smoothing and edge positioning. The combination of canny operator and the method of edge detection based on dyadic wavelet transform, a method has been proposed on the basis of edge image fusion, many experiments show that this method is better.