| Abstract: | Transmission of images over low speed, low bandwidth channels has several applications. Examples include transmission of remote sensing images, scanning of vast global databases and facsimile transmission of printed material. In progressive image transmission, a low resolution image is first transmitted with as few bits as possible. The resolution of the transmitted image is then improved interactively upon viewer's request. Eventually, an exact copy of the original image is reconstructed. In this thesis, image coding techniques and parallel architectures are first reviewed followed by a review of progressive image transmission techniques. A software toolkit for progressive image transmission is then presented. This tool is user-friendly and permits rapid development of new algorithms and fast evaluations of different approaches for progressive image transmission. Unlike conventional techniques, it offers a convenient platform for performing both objective and subjective evaluations with ease. Two algorithms for progressive image transmission are then detailed. The first algorithm follows the coefficient scanning methodology. The second algorithm uses an iterative residual error feedback technique. An architecture which implements these algorithms in real-time is then presented. The software tool is used as the platform with which the comparative analysis between different algorithms are performed. Simulation results are reported for the proposed algorithms in terms of Peak Signal to Noise Ratio (PSNR). Simulation results indicate that the proposed algorithms perform better than JPEG-DCT in terms of subjective and objective quality. However, Gabor decomposition is computationally expensive and hence requires special purpose architectures for real-time implementation. The proposed architecture achieves a high efficiency of parallelism (99%) and implements the two algorithms for progressive image transmission in real-time. The architecture is simple and modular and hence can be easily implemented in VLSI as a codec. |