An image edge can be defined as the difference of image features in a local region, and its appearance is the mutation of image g ray or texture structure or color. Image edges are very important to human being and machine vision, because they can transfer the most information of an image. Detecting image edge is considered as a key step in many complicated processing methods like image segmentation, image recognition and feature extraction. In this paper, some classic methods are discussed and the à trous wavelet decom position applied to detecting image edge is discussed based on the wavelet decomposition theory in detail. The à trous decomposition is one of discrete wavelet transform algorithms. On the basis of à trous wavelet decomposition theory a detecting image edge method is derived, according to which the image can be decomposed into wavelet planes of increasing scales and the wavelet planes have the same number of pixels as the original image. An algorithm of detecting image edge based on à trous wavelet decomposition, which is suitable for computer program is presented. We use the method, Sobel and Robert algorithms to process the same SPOT image. Comparing the results, we can find that the m ain edges which are detected by the à trous wavelet decom position method from the origin SPOT image are better than those obtained by Sobel or Robert method. We also note that the new detecting image edge algorithm can gain more tiny edge information than the other two methods. When the original image is stained by noise, the detecting image edge algorithm on basis of the à trous wavelet decomposition is not disturbed mostly, but Sobel and Robert algorithms are sensitive to noise. We have applied the algorithm which is presented for detecting image edge to getting binary image, and the result is satisfactory. So, the algorithm is better than the other two methods in some aspects, especially applied to original images which are stained by noise.