Abstract: Even though the wavelet based image compression algorithms have been well(recognized) recently,the compression results of these algorithms are not well enough for the images with rich textures,since the spatial pattern characters of an image which are important for human vision have not yet been involved in these algorithms directly.A new image compression method based directly on pattern characters of image named PCBC(Pattern Character Based Image(Compression) Method) is introduced.The results of testing on a set of remote sensing images with rich textures show that the PSNR(Power Signal-to-Noise Ratio) of PCBC is higher than that of SPIHT(Set Partitioning in Hierarchical Trees).