With the novel idea of describing an image in characteristic subspace@CS), a new approach of the more efficient image compression for some sorts of images, such as medical sciagraph, is proposed in this papers . Firstly, we are to obtain the coefficient image of the original image by its description in DCS; then decompose the coefficient image with forward wavelet transform ( WT 1 in multi-resolution to remove the redundant information; and encode the decomposed image with vector quantization(VQ) at last. In comparison with WT-VQ or DCT-VQ, the new way@CS-WT-VQ) given by us has been proved by corresponding experiments that DCS-WT-VQ has much higher compression ratio, SNR and less computational time in coding and decoding process.