Adaptive image compression based on regions of interest and a modified contrast sensitivity function

An algorithm for still image compression is presented; it is based on wavelet transforms and embeds a model of the human visual system. The novelty of this approach is the introduction of three main features: (a) the algorithm allows a user-oriented definition of multiple regions of interest in the image, where each region can be associated to different visual errors, independently of the others; (b) a model of the human visual system is embedded in the compression task by preserving the spatial information in an adaptive scheme of quantization; and (c) the adaptive quantization, based on a modified contrast sensitivity function, is integrated in the classical embedded zero tree coding to better exploit the multiresolution formulation of wavelets. Experimental results are shown to validate the theoretical assumptions and to compare this solution with other approaches