Image coding with a map and vector quantizer
C. Lo · IEEE International Conference on Acoustics Speech and Signal Processing · 1993
An image coding scheme with low complexity and good performance is proposed. A map is used to split the image into two parts, where part A consists of blocks which are hard to code but bears most of the visual information, and part B consists of blocks which are easy to code but bears little visual information. The coefficients taken from a 2-D discrete cosine transform are used to separate the two parts and to generate the map. After the map is created, one uses two LBG vector quantizers to select the codebooks for each part and sends the addresses as well as the codebooks. The algorithm has been applied to the standard 512*512 Lena image, and an MSE (mean square error) of 61.82 (peak SNR of 30.22 dB) at a compression ratio of 21.40 (0.3738 bit/pixel) has been obtained.>