Context-Based Predictor Blending for Lossless Color Image Compression

Abstract

Images are typically nonstationary signals. If prediction is applied in a linear fashion, it must be combined with a technique that takes this characteristic into account. In general, images can either be regarded as piecewise 2-D autoregressive processes or they are handled in a blockwise manner. This paper presents a novel prediction technique, which treats the image data as an interleaved sequence generated by multiple sources. The challenge is to deinterleave the sequence and to compute prediction weights for each subsource separately. The proposed approach adaptively determines the subsources based on the textures surrounding the pixels. The new linear prediction technique is combined with template-matching prediction and a blending method that considers the correlation between the predictors’ estimates is proposed. The prediction method is incorporated in a framework for lossless color image compression. In combination with an adaptive color transform and a dedicated coding algorithm, the proposed approach shows a competitive compression performance for a wide range of natural color images.