A new algorithm for image enhancement based on the dyadic wavelet transform is proposed aiming at minimizing the image noise while amplifying the image feature. By making full use of the shift invariability of the dyadic wavelet transform and correlation of the wavelets coefficients from different scales, the proposed algorithm can efficiently alleviate the amplification of the noise and distortion of edges, which are the most difficult and common problems in image enhancement. Furthermore, the enhancing strategy is highly adaptive and thus applicable to images from many sources. The results of experiments are also given, which show that the new algorithm performs better in both de noising and image feature enhancing compared with the existing multi scale algorithms.