An optimization based approach to color image compression using a Rate-Distortion model

In this work a new model for color image compression is introduced. The model predicts the expected distortion in compressed images as a function of the compression rate, known as a Rate-Distortion curve. Based on the model, we present anon-linear optimization problem, from which the optimal color components and rate allocation are derived. We extend the problem to consider down-sampling of the color components. We show that the rate-distortion model in conjunction with the probability distribution of subband coefficients can be used to develop an efficient algorithm for coding color images. We demonstrate this approach for the Discrete Cosine Transform (DCT) and the Laplacian distribution as the probability model. We measure the distortion using Mean Square Error (MSE) and Weighted Mean Square Error (WMSE). Simulation results are presented and discussed to support the efficiency of the new algorithm.