Method Of Image Edge Detection

The edge of image is often considered as its fundamental feature in image processing. It has been used in image processing and analyzing technologies such as feature description, image recognition, image segmentation, image enhancement and image data compression on a high level. Then we can comprehend and analyze the edge image later on.The point of singular signal's conjoint pixel always changes acutely in its gray-level value. The gray-level distributing gradient of conjoint pixels can show this change. After introducing the conventional edge detection operator and multiscale wavelet edge detection operator, we discussed the well quality of B-spline function > n-class derivative of Gauss function N Harmonic function and Hermite function in wavelet theory and their concrete application in the image edge detection.We may study the Lipschitz exponent characterization of the noise and singular signal and then achieve the goal of removing noise and distilling the real edge lines. The thesis has discussed the calculating of the Lipschitz exponent, and analysised and compared the condition between wavelet bases and singularity detection of signal.In addition, the thesis has discussed the technologies of the clearing image and enhancing the contrast before the edge detection, and study the problem of the threshold selection and the edge line connect processing for gained the better edge image.Finally, this thesis discussed these following questions: first, the algorithm of used the error image for improving the purpose of the edge detection. Secondly, we have transformed the solved question of the first and the second directional derivative to frequency domain and founded they have a single formulae in frequency domain. Thirdly, we have described the Singular signal and the noise by using the correlations of the neighbor data after wavelet transform. At last, the edge detection is turned into the minimization of the total energy through making a new energy function and adaptively adjusting the weighting parameter.