Multiscale Edge Detection Analysis Bases On The Wavelet Transform

Object edge usually exists between background and object, between object and object, area and area. It draws the outline of the geometric object and impresses much information of image and describes the important characters of the vision of object and offers valuable and important parameters for people to describe or identify object or interpret images. This information pays a direct and important influence for us to proceeding high degree image process, such as image filters, character description and pattern identify. Therefore, edge detection shows an important and key position in image process.There are many scholars who have contributed in the domain of edge detection of image in the world since people put forward the conception of edge detection of image in 1965 years. Usually, the classical edge detection algorithms base on the relation of derivation of the image pixels to proceeding edge detection. There are some classical edge detection algorithms base on one order or second order of image pixel to detection edge, such as the Roberts algorithm, the Prewitt algorithm, the Laplacian algorithm, the Sobel algorithm, the Canny algorithm, and so on. Commonly, these algorithms base on square template in the digital image process. But these edge detection algorithms detection edges only under a scale context, they can't well detect the local change of image.The peculiarity of multiple resolution analysis of The Wavelet analysis affords a new method for edge detection; we can extract the local character of signal by analyzing the signal with multiple resolution analysis of The Wavelet Transform and pick up the edge of images at the same time restrain effectively the noise. As a result, the Wavelet function possesses stronger competence of wiping off noise and competence of multiscale edge detection at well.Presently, multiscale edge detection is a comparative novel topic in image process and attracts many scholars to make great efforts for it. People detect edge of image and deal with the result of edge detection under different scale by the multiscale edge detection algorithm and synthesis the result according to different needs. People can get a image process result which well cater to the require of people by synthesizing the information which people get in each scale.The principal axis of this paper is the multiscale edge detection which base on the Wavelet transform. This paper introduces briefly the basis theory of the Wavelet transform and image process and introduces briefly Wavelet transform single scale edge detection. Following, this paper introduces the important contents——the Wavelet transform multiscale edge detection algorithm. This paper makes use of the multilayer details of two dimensions image wavelet decompose to construct three edge detection methods in a creative way. The first multiscale edge detection method bases on the wavelet decompose details. The second multiscale module maximum edge detection method bases on the wavelet decompose details. The third multiscale edge detection method bases on the wavelet decompose details module maximum and data amalgamation. The three methods are a gradually approach relation, the second algorithm deduce from the first method, the third algorithm deduce from the second method. In this way it produce a interlink knot role in thinking. Furthermore, this paper demonstrates the result of edge detection which bases on the background of the standard Lena image by these three methods and compares the detection result with the detection result by the classical edge detection algorithm——Canny algorithm. We can find that the detection result by the three methods has different characteristic from the analysis in the paper. The three methods have different roles to the change of image pixels and produce different edge detection results. Although, the wavelet transform multiscale edge detection which was introduced by this paper can detect more details. It has some shortcoming of detect ability in area of the intension relative smoothly changes.