Research Of Image Edge Detection Based On Wavelet Transform

Many fields related to image analysis are based on the edge detection of digital images, such as image segmentation, recognition of target region, shape extraction from regional areas and so on. The edge detection of digital image is an important attribute for extracting image characteristics in image recognition. It is also the first step in image understanding and image analysis. It has become one of the most active issues in the field related to machine vision research. In this thesis, the present situation in Image Edge Detection Technology research and the development of wavelet transform are analyzed; meanwhile, the research significance of Edge Detection Technology based on wavelet transforma is elaborated. By the analysis of the current Image Edge Detection Technology and according to the contradiction of noise control and edge detail extraction, Edge Detection Algorithm based on wavelet transform is proposed to carry out image edge detection.With the development of Image Detection Technology, different approaches of edge extraction have been applied widely in the fields such as engineering, industrial production, agriculture, military affairs, medicine and scientific research. These approaches improve the performance of edge detection and have good application prospects. For example, the noise reduction of medical imaging, quality inspection of agricultural products, such as the appearance.Despite all this, the edge extraction of digital image has not been fully solved: the decrease in edge gray-scale variation is accompanied by the difficulty in edge extraction; the noise generated during the production and transformation of the images may result in false edge and undetected edge; owing to the restriction of snapping environment and condition, there may be some irrelevant interferences. It is the major challenge in image processing to improve the accuracy of edge detection and the signal-to-noise ratio of edge extraction algorithm, thus making the algorithm an emphasis of professional study. Also we have to work towards it. Multiresolution analysis theory and Binary Wavelet Fast Algorithm are introduced in the thesis. Through the analysis on Signal Singularity Detection in wavelet transform, wavelet transform is proved to be efficient in detecting deformed signals. According to wavelet basis function selection criterion in edge detection and Canny optimum edge detection principle, B-Spline is chosen as the wavelet basis function of edge detection. B-Spline is constructed in Multiresolution Analysis Frame. The spline space on the basis of B-Spline forms the Multi-scale Approximation of signal space then first and second derivatives of the approximation signals of different resolutions are evaluated and module extremum detection or zero-crossing detection is done respectively, thus the multi-scale wavelet detection operators of B-Spline is formed. Cubic B-Spline function is chosen as the wavelet edge detection operators because module extremum detection exceeds zero-crossing detection.This approach excels in avoiding of noise interference but some weak edges may also be missed as well. In order to compensate for this deficiency, an improved B-Spline wavelet edge detection algorithm with adaptive threshold is proposed: on the basis of edge detection algorithm of multi-scale B-Spline wavelet, adaptive smooth filter is adapted to sharpen the image edge to prevent the undetected weak edge. According to the computer simulation, the modified B-Spline wavelet edge detection algorithm using adaptive threshold maintains the weak edge together with noise elimination.