Techniques for image enhancement and robust video communications.

摘要

With the improvement in communication technology, broadband services are readily available for general multimedia communications. While the bottleneck in communication bandwidth has been relieved, the quality of the delivered images and videos becomes the next major concern when designing the communication systems. In this research work, we consider two important topics as far as quality image and video communications is concerned: (1) image enhancement; (2) source coding schemes for reliable transmission.; Current studies on image and video communications consider only the problem due to the transmission channel. To achieve quality image and video delivery, the problems incurred when acquiring the images or videos should not be ignored. In this work, we investigated two major kinds of image or video acquisition problems: blurring and noise contamination. We study the enhancement techniques for such problems. In particular, to deal with the blurring problem of acquired images, a fast algorithm is proposed for blind image restoration based on the discrete periodic Radon transform (DPRT). With the DPRT, the original 2-dimensional blind image restoration problem is converted into 1-dimensional ones such that the complexity is much reduced and thus faster restoration can be achieved. Besides, to enhance the performance of compression with noisy acquired image, we investigate the adaptive thresholding technique that is particularly suitable for the MPEG-4 still texture image codec in coding noisy images. It integrates the denoising process and the coding process into a single step for the traditional MPEG-4 still image coder such that the visual quality of the decoded images are much improved.; For video communication, a novel model is proposed to describe the relationship between the quantization and the rate-distortion (RD) performance of block-based video codec. We show that the distortion-quantization relationship, D(Q), of most coded sequences, encoded with different group-of-picture structures, can be approximated by a single linear model using regression analysis. We further verify the rate-quantization relationship, R(Q), based on the proposed D( Q) model. Estimates from the proposed models are compared with actual measurements and promising results are obtained. Furthermore, to realize robust transmission, a new error resilient transcoding scheme for H.263 video transmission over combined wireless and wired channel is proposed. It first determines the coding type of every group-of-block in each frame during the coding process by considering the RD performance of the coder. A transcoder is then proposed to assay the transmitted bitsteam and revise the corrupted parts in case of error. Finally, the revised bitstream is sent to the decoder through a reliable channel. As compared with the traditional error resilient approaches, the proposed scheme greatly improves the quality of the decoded video.