Wavelet Transform In The Study Of Edge Detection Of Super Vacuole Image

There are much information in the edge of the image,through the edge detection,the edge of object can be extracted from the image,and the geometry contour information also can be obtained, this information is important for the object recognition and feature measurement.At present,the edge detection occupies a key position in image processing and machine vision,which is the important field of image segmentation, recognition research and application.Although there are many theoretical research findings in the edge detection of image in these days,the each algorithm still has many problems in specific application.The traditional edge detection algorithms are mainly based on a differential operator, the typical measures are Roberts,Prewitt,Sobel,Kirsch,Canny,Log,the advantages are less calculation amounts and short execution times,but its drawback is very sensitive to noise. In this paper, the edge be detected from the image of super vacuole, the image is obtained from the underwater with the high peed projectile with much noise,the quality of image be affected by underwater environment and ligh, how to obtain the valid edges from the noise image is the focus of this paper.Wavelet transform can multi-scale approximate the image with good local and multi-scale resolution features.On small scale,the result contains much detail edge,on large scale, many weak edges and noise be filtered out, leaving relatively stable edges.Synthesizing the multi-scale edges can overcome the noise and preserve the weak edges;the complete edges are effectively extracted.The wavelet transform is applied to the edge of the super vacuole image,the adaptive multi-scale wavelet edge detection algorithm is proposed in this paper,the algorithm use anisotropic diffusion filter for denoising and enhancement,then select the Gauss or B-spline function as the base of wavelet, through the wavelet transform obtain many scales of edge information,it use the 5 kinds of adaptive threshold methods to detect the each scale edge, comparing the efficiency and robustness of the methods,it use the K-mean threshold scheme as the better means, the each scale edges are fused base on the two different fusion schemes to get the final edge.The ultimately edge of the super vacuole image prepared the ground for shape feature extraction and 3D reconstruction.This edge detection draw more accurate effective and intact edge of Super Vacuole than the traditional mehod.