Wavelet-Fuzzy-Stochastic Kalman Filtering for Image Compression

This paper presents a novel fuzzy stochastic Kalman filter for compression of digital images. In particular, it is shown that the state evolution of the synthesis coefficients of any discrete wavelet transform (DWT), in presence of coding degradation, may be described fuzzily. The novelty of this description is that, unlike other fuzzy based methods, it does not require a predefined membership measure. The fuzzy representation is further characterized by a stochastic nominal value and an interval of uncertainty. Furthermore, traditional DCT based coding is judicially applied to the smooth regions of the DWT. It is shown that such a framework allows for an efficient coding of images