Abstarct: Increase in image size caused image compression to become a major subject is image processing. Hence, improving image compression techniques is considered an important issue. Image compression is performed using redundancy in the image. A higher data redundancy in an image can lead to a higher compression ratio in image processing techniques. Although the compression ratio of an image processing technique can significantly be improved using pre-processing technique, this approach has less been considered by the researchers. Although lossy compression methods, in contrast to lossless methods that can exactly recover the original image, have a high potential in image compression, they produce distortion in the compressed image. In implementation of an image compression algorithm, knowledge about the function of the human visual system is important. The sensitivity of the human visual system is not alike on all areas of an image, indeed some information in an image cannot seen by the human visual system. In this thesis, by utilizing the saliency map, in which the importance of various areas is determined according to the human visual system, an adaptive image compression is performed. Indeed, a higher compression ratio is considered for areas with a greater sensitivity. Because of the importance of accuracy in extracting the salience map for adaptive image compression, in this thesis, a new method is proposed for extracting the image salience map which outperforms the other existing methods in extracting the image salience map. Although the adaptive image compression reduces the distortion in lossy image compression methods, distortion is yet unavoidable. In this thesis, several post-processing methods are proposed to reduce the effects of distortion due to the image compression methods. Since the reference image and information about the amount of distortion are not available for post-processing, the quality of compressed image should initially be evaluated without having the reference image. Then, according to the image quality, an appropriate post-processing is applied on the compressed image. Since the efficiency of image quality assessment has an important role in the performance of image post-processing, a new method, superior to the existing methods is proposed in this thesis for no-reference image quality assessment of JPEG compressed images. To evaluate the proposed methods in the thesis, several data-baxses such as CSIQ, LIVE II, IVC, Kodak and MSRA are used. Although the pre-processing and post-processing techniques can improve the compression ratio per se, the simultaneous use of them can further increase the improvement. Indeed, by the simultaneous use of these two approaches, in addition to benefit from their advantages, a possible distortion due to pre-processing can be removed in post-processing step.