Abstract: Because the wavelet transform has good local performance in both spatial and frequency, it is especially suitable for edge detection in images.For resolving the conflict between denoising and local accuracy,the paper suggesed a new edge extraction method of the power correlation of different directional wavelet transform at several adjacent scales in terms of the different correlation of signal and noise wavelet transform contents.Vertical and horizonal edges in images first were enhanced,then the modulus maximum valuewas evaluated.Under denoising effectively,the detector can preserve accurate localization.That was also proved by final experimental results.