DATA-COMPRESSION FOR IMAGE ARCHIVING APPLICATIONS

A novel approach to still-picture data compression is presented that i s based on the integration of two different coding techniques. Its aim is to obtain high-compression factors by exploiting the characteristi cs of a second-generation method (i.e., polynomial approximation) in s uitable scene areas, while achieving higher fidelity, where required b y the local image characteristics, through a first-generation techniqu e (i.e., vector quantization). Such a region-driven hybrid approach pe rforms the integration by utilizing a control image (resulting from th e detection of areas characterized by high spatial frequencies) that d rives the segmentation and approximation processes. The results obtain ed on a variety of natural color images are quite satisfactory. The vi sual quality of the reconstructed images is, in general, very good, an d SNR is kept within an acceptable range. Moreover, the bit rate obtai ned is always below 0.15 bit/pixel (bit/pixel), with peak values of 0. 07 bit/pixel.