Because of cloud in remote sensing images, the images were fuzzy and the information of earth objects could not been read. A new method for removing cloud had been projected based on wavelet analysis. The 1～8 orders decomposition for several gray remote sensing images 256 pixel×256 pixel was implemented with wavelet analysis and the detail coefficients were multiplied with 1～10 times in the process of reconstruction. Firstly, the order 7 of decomposition and the times of detail coefficiency had been selected by maxium entropy with selected wavelet function. Secondly, the functions of bi-orthogonal 2.2 wavelet and discrete approximation of Mery wavelet had been found by maxium entropy with selected order number and times. Finally, the exact order number and times had been shown in the curves of 1～8 orders decomposition and the times from 1 to 10 with 0.1 step length. The results show that the times 1.1 was best times for removing cloud from remote sensing images with thin cloud, and for discrete approximation of Mery wavelet there were two peaks in the times as 4.8 and 5.7.