Removal of subjective redundancy by adaptive quantization of DCT coefficients

The removal of subjective redundancy in image coding has been playing a very important role in studies of the low bit rate video communications to attain even greater levels of signal compression. This paper discusses in depth the psychovisual interactions between the basis pictures (components) of DCT in the transform domain, and based on this, suggests a coding scheme to further reduce the required bit rate by adaptively quantizing the DCT coefficients. The general goal is to maintain the perception of reconstructed images as good as possible, therefore, several human visual system's characteristics such as the responses to intensity and to spacial frequency as well as the spacial masking effect are fully exploited. Some experiments are carried out, and very encouraging results are obtained.<>