A vector coding with a compressive algorithm for natural images

This paper proposes a new vector coding with a compressive algorithm for natural images. Here the pixel-value-boundary contours on RGB planes of natural images are expressed by function approximation, and their contour shapes are predicted from similar contours in the decoder side. A vector coding, which represents images by function approximation, not only meets the demands of expansion and reduction but also suits content distribution because of the keeping quality of images even for the repetition of editing. In spite of these merits on vector coding (representation by function approximation based on characteristics of images), the application to natural images has never been studied owing to its considerable increase of data. To put vector coding for natural images to practical use, we try to compress the data by using characteristics in three-dimensional representation of images