Professor of electrical and computer engineering, beckman institute
Video and images are important components of communication. With the growing use of multimedia and the Internet, efficient and easy-to-interpret video and image representations have become key technologies. New methods are needed that can provide good-quality video and images with resources that are limited in computational power and bandwidth. This work is aimed at developing a framework to inculcate into visual communication methodology the use of the spatiotemporal structure present in the data, with the goal of increasing both the efficiency and perceptual quality of the communication. Commonly used techniques have proven inadequate in providing satisfactory performance in many situations of practical interest. But Ahuja recently has developed an approach that yields a hierarchical description of the multi-scale structure in images and video. The hierarchy captures perceived regions of all shapes, sizes, and contrasts at all a priori unknown scales. During his Center appointment, Ahuja will formulate a general methodology that combines the new hierarchical description capability with traditional techniques in computationally efficient frameworks. Four scenarios will be used as the foci: compression of still images with minimal loss in perceptual quality, restoration of noisy images with minimal perceptual loss, magnification of the image/video size with minimal artifacts, and compression of video with minimal loss of perceptual quality.