ABSTRACT
Quantization is essentially discretization in magnitude, which is an important step in the lossy compression of digital image and video. The input and output of quantization can be either scalars or vectors. Uniform quantization is the simplest, yet very popular quantization technique. Conceptually, it is of great importance. The companding technique, also known as logarithmic quantization, consists of the following three stages: compressing, uniform quantization, and expanding. Adaptive quantization attempts to make the quantizer design adapt to the varying input statistics to achieve better performance. There are two different types of adaptive quantization: forward and backward adaptations. The selection of block size is a critical issue. If the size is small, the adaptation to the local statistics will be effective, but the side information needs to be sent frequently. Pulse code modulation (PCM) is closely related to quantization. PCM is the most important form of pulse modulation.
