A Study of Bias Correction Methods for Enhancing Median Edge Detector Prediction

Abstract

In this paper, we present three novel lossless compression approaches for gray-scale continuous tone natural image. Our methods enhance the median edge detector (MED), which is the core part of JPED-LS algorithm, by reducing the entropy of the prediction error via adaptive regression. These modified predictors improve the prediction accuracy by reducing the negative effect due to MED's oversimplified edge orientation detection. The experimental results show that our approaches achieve evidently better performance than MED with only neglectable increasing of computational complexity and without introduce extra pixels into the causal template