We proposed a new image compression method based on region of interest, in order to reduce the storage space of image and lower the distortion of image. The eye tracker was used to extract image regions of interest, and then partition compression of image was conducted by making a compression mask. The background area was compressed using DCT algorithm, while the area of interest was kept as original without any compression. Subjective and objective evaluation results showed that the compressed images had little distortion and good observation effect, and the storage space was reduced by half. The method was simple and the compression efficiency was high. In conclusion, the compression method based on extracting region of interest through eye tracker was superior to the compression method based on Itti visual model. This method has a good practicality and can be used to compress many types of images.