JPEG image compression algorithm (JPEG-HVS) was studied based on improved quantization table combined with characteristics of Human Vision System. It calculated a new kind of quantization table according to the human visual luminance contrast sensitivity function and used it to take place of the luminance quantization table form JPEG standard, and carried out the simulation experiment on different classes of images via Matlab7.0. By comparison, the JPEG-HVS had a better compression ratio, 53.56% higher than the traditional JPEG compression algorithm and 18.75% higher than the JPEG zone method. The difference of peak-signal-to-noise ratio (PSNR) among the three was very small, i.e. JPEG>JPEG-HVS>JPEG zone method. The mean structural similarity index measure (MSSIM) of the three was JPEG>JPEG-HVS>JPEG zone method. The time of coding and decoding of JPEG-HVS was far less than that of JPEG. Meanwhile, images decompressed by JPEG-HVS still had a good visual effect by observation. Compared with the other two, with the quality of compression image guaranteed at the same degree, the JPEG-HVS can reach higher compression ratio and faster coding and decoding, which will benefit the storage and transmission of images.