It will be a great convenience to compress images on a high proportion with the retention of details for transfering, storing, and processing images. In view of that, a new pretreatment technique - wavelet coefficients partition (WCP) algorithm was presented in this paper, which can subdivide the wavelet decomposing high frequency coefficient into high-frequency information and low-frequency information by setting threshold and iterative calculation. The low-frequency information can be compressed by 9/7 wavelet based on the promotion way. On the other hand, the detail information of the high frequency can be processed by the second generation Bandelet. Meanwhile, the specific optimization algorithm was also given in this article, through which the calculation can be reduced by three-quarters; and comparing with the original image, the PSNR value of the same bit rate decompressed image can be improved to the maximum of 10 db, the SSIM value tends to be 1. The experimental results show that the compression method is easy to run, and the visual effect of the restoring image is better than any of those that have been compressed by wavelet transform, JEPG2000, Bandelets and other conventional methods.