Lossless compression of images employing a linear IIR model

An algorithm for the lossless compression of different classes of images is proposed. This approach is based on modeling the original image by a rational function which consists of poles and zeros, or equivalently an Auto-Regressive Moving-Average (ARMA) process. The equation-error structure, which approximates the image by minimizing the error in the least square sense, is used to obtain the optimal coefficients of the transfer function. This technique is implemented in the frequency domain. The performance of the proposed approach for the lossless compression of different classes of images is evaluated and compared with the lossless linear predictor. The residual sequence of these schemes is coded using arithmetic coding. The suggested approach yields compression measures, in terms of bits per pixel, lower than the lossless linear predictor for compressing 8-bit gray-scale images