An image coding scheme based on image projections and geometrical VQ

摘要

This paper deals with an image coding scheme based on overlapped projections and geometrical vector quantization. As far as still images are concerned, we present a simple coding algorithm, in which, the concept of projection, normally employed in tomographic applications, and geometric vector quantization are used to describe, in a compact way, circular overlapping regions in which the input image is subdivided. In this way it is possible to represent in a good way, contour and edge information without introducing significant blocking artefacts. The proposed coding techniques can also be used to represent images at progressive levels of quality. This aim can be simply obtained by increasing the number of projections used to describe each image region. Simulations carried out on several test images showed interesting results. An extension of this projection-based coding techniques is proposed for moving images too. In this case projections are used to describe in a compact way, the motion compensated luminance differences between the current image and the previous coded one.