Abstract: The paper proposed an image edge detection algorithm hased on multi-direetional and multi-scale wavelet transform for the shortcomings of the poor ahility of anti-noise and edge locating accuracy of traditional edge detection algorithms.The algorithm first take wavelet transform to the image from multi-directions and with multiscales, then the maximum wavelet coefficients in all directions and various scales of the image arc fused by EM algorithm,in the end,the image edge acquired through a maximum entropy threshold processing procedure. Experimental results show that the algorithm is more computational efficient for it omitted the process of evaluate the wavelet module and more precisely edge localization by search the various image direction as possible by multidirectional wavelet transformation.