Efficient color passport size photograph compression using ROI coding

Imaging applications require processing, transmission, and storage of enormous digitized data produced from large sets of images. To conserve storage, increase the processing speed and to make their transmission faster on low bandwidth channels, images are compressed prior to their storage, transmission, and processing. Different approaches to compression have been developed to achieve desired degrees of compression ratio. Techniques that achieve highest degree of compression may reduce the size of image considerably but may also reduce the quality of image substantially. In certain types of images such as passport size photographs, loss of quality in the region of interest is highly undesirable; however, some loss of information in non-regions of interest can be accepted. For such images region of interest can be left uncompressed or compressed through lossless or near lossless algorithms. To achieve good quality of reconstruction of region of interest and better overall compression ratio, this paper proposes the use of multiple compression techniques for passport size photographs. It investigates through Matlab the application of different lossless compression techniques in terms of peak signal to noise ratio, compression ratio, and mean square error on region of interest together with lossy compression techniques on non-region of interests of passport size photographs to determine best possible compression technique and compression rations for both regions.