A hybrid image coder with vector quantizer

摘要

In this paper the scheme of a hybrid coder is presented. It uses a motion compensated predictor to obtain, from the present image and the previously coded one, a new image which represents the motion compensated luminance differences. Then, this new image is vector quantized. Each vector contains 16 elements (obtained from a square block of 4 × 4 pels). Four different codebooks are used for the quantization, taking into account the local image detail. To reconstruct the images at the receiver side, the estimated motion field must be transmittes; this is done using certain adequate compression techniques. At about 0.5 bpp, the coded images, obtained by proper simulations, have good quality, close to that obtained with a hybrid DCT coder.