Variable-rate contour-based interpolative vector quantization for image coding

In order to obtain perceptually enhanced interpolation, the authors propose a directional technique that follows edge contours to interpolate pixels along and near contours more accurately. This technique, referred to as variable rate contour based interpolative vector quantization (VQ), uses the directional information inherent in local gray level transitions and consists of an orientation estimator followed by directional interpolators. By combining a powerful VQ technique with decimation and directional interpolation, a very efficient image representation is obtained. An adaptive corrector stage using variable-rate VQ is added as a final stage of coding to improve the quality of the reconstructed image. Some relatively large amplitude errors tend to persist along image edges, and the adaptive scheme helps to correct the remaining errors in those regions selectively using an edge detector. Contour-based interpolative VQ yields fairly good reconstructed images at about 0.27 bit/pixel with a relatively simple vector coder structure