In order to store and transmit largely and efficiently, SAR images must be compressed effectually. The traditional coding method of SAR image based on wavelet translation can only be used to decompose low frequency, so a lot of information of intermediate/high frequency is lost. In view of the above problem, this paper presents an adaptive SAR image compression algorithm based on wavelet transform. In this paper, the speckle noise is reduced to improve the image quality first. And then according to the energy index, the SAR image is decomposed into a series of subbands, which have different significances. Finally, all the subbands are quantized by the minimal error quantization at the same bit rate. The use of the algorithm can protect the high frequency details of SAR image adequately, and can achieve a better compression result.