Wavelet-based Entropy Measure for Rate-Distortion Optimization in Image Coding

A novel method for calculation of the entropy measure in wavelet space is proposed. This perceived-based entropy measure uses a Second Order Model entropy estimator, in which the occurrence of neighbors is considered in formulation. It has the intention to allow the implementation of a more suitable measure in coding processes and a relationship between the metric and the description of perceptual features. This method is used for the Rate-Distortion optimization in order to improve the bit-allocation coding algorithm, demonstrating that the wavelet-based entropy estimates a truncation step close to the target rate. The hypothesis is founded in the effect of distortion on the coefficient allocation. Because entropy measure is a close approximation of the conditional probability of image in multi-resolution space, it provides an adequate representation for the information of a Detail feature. A definition of Detail-based homogeneity variance criteria is used for the information quantity - wavelet representation space, in order to find the image that fits a given Quality Level criteria. Experimental results are obtained for artificial and natural databases.