Lossless color image compression using chromatic correlation

[Summary form only given] Typically, the lossless compression of color images is achieved by separately compressing the three RGB monochromatic image components. The proposed method takes into account the fact that high spatial correlations exist not only within each monochromatic frame but also between similar spatial locations in adjacent monochromatic frames. Based on the observation that the prediction errors produced by the JPEG predictor in each RGB monochromatic frame present very similar structures, we propose two new chromatic predictors, called chromatic differential predictor (CDP) and classified CDP (CCDP), to capture the spectral dependencies between the monochromatic frames. In addition to prediction schemes, we consider context modeling schemes that take into account the prediction errors in spatially and/or spectrally adjacent pixels in order to efficiently encode the prediction errors. In order to demonstrate the advantage of the proposed lossless color image compression scheme, 5 different types of images are selected from the KODAK image set. All images are RGB 24 bpp color images with resolution 768×512. The experimental results demonstrate significant improvement in compression performance. Its fast implementation and high compression ratio may be a promising approach for the application of real-time color video compression