Predictive vector quantization with ridge regression

Prediction can play an important role in image compression. Better rate-distortion tradeoffs can be achieved by coding the residuals from predictive schemes rather than the direct pixel values. The price is only a modest increase in complexity. A variety of linear and nonlinear predictors have been used successfully in applications of predictive vector quantization. It makes sense that the better the prediction, the better the resulting compression. One method by which to improve predictive accuracy is to reduce the variability of the predictions. We here apply ridge regression in order to obtain prediction coefficients for use in a predictive vector quantizer as an alternative to standard Wiener-Hopf techniques