
91Image Compression Using Vector Quantization
simple VQ and gain-shaped VQ when coded at 0.25 bpp with a 4 × 4 block
size (Bovik 2005).
5.7 Classified Vector Quantizer
As discussed in Section 5.2.1, the major drawback of VQ-based compression
is blockiness in decompressed images. Edge information is lost or degraded
in small-size codebooks. In order to preserve the edges, it is necessary to
have such codevectors present in the codebook. Gersho and Ramamurthi
(1984) worked on this problem by classifying each image block vector into
different classes or categories. Some of the classes are as follows:
1. Shade with smooth variation of gray levels ...