There is a considerable amount of redundant information in image data which makes image compression possible. Redundancy of data

spatial redundancy in particular varies with different images. It is necessary to study the spatial redundancy of compressed images and reduce the random selection of image compression methods. In this paper

a novel idea of image classification for image compression is proposed and its algorithm is presented too. The distribution of wavelet high frequency coefficients in images is considered while edge active measure (EAM) is defined to describe the nature of images in this algorithm. By EAM images can be classified and compression result can be forecasted .The experiments have shown that the image classification and result forecast implemented in this paper make sense and correspond to human visual understanding. The idea suggested in this paper has been of great value to the election and optimization of algorithms for different purposes.