Nonhomogeneous image coding for progressive transmission

A new progressive image transmission scheme which uses eye perception is described. The algorithm applies a very fast quantization technique to images represented in a pyramidal form. The pyramidal data structure has relatively low equivalent entropy and is suitable for progressive transmission. To gain further compression a very fast vector quantization is applied to the pyramidal representation of the images on a level-by-level basis. This fast vector quantization technique is based on the minimax error criterion, such that the quantization error will be within a prespecified bound. The quantization error at each level will be delivered to the next level up to the bottom layer wherever required to ensure that the user-defined more important areas will be assigned a small allowable quantization error, while maintaining the rest of the image within a slightly higher prespecified bound. Simulation results showed the use of nonhomogeneous image coding to obtain comparable quality with a lower bit rate than for other complex techniques. >