Least squares based optimal switched predictors for lossless compression of images

Gradient Adjusted Predictor (GAP), used in CALIC, consists of seven slope bins and one predictor each is associated with these bins. As the relationship between the predicted pixels and their contexts are complex, these predictors may not be appropriate for prediction of the pixels belonging to the respective slope bins. In this work, we present the Least-Squares (LS) based approach to find optimal predictors for pixels belonging to various slope bins of GAP. Our simulation results show that the proposed method results in similar performance as that of Edge Directed Prediction (EDP) and Run-length and Adaptive Linear Predictive (RALP) coding. EDP and RALP use symmetrical encoder and decoder structure. On the other hand, we propose an unsymmetrical codec that has higher encoding complexity but decoder is very fast - as fast as a decoder based on GAP principle. However, our encoder is computationally much simpler than an EDP and RALP based encoders.