Efficient lossless compression of colour image data with diverse characteristics

Lossless compression of data is the attempt to model the data in such a manner that the resulting file contains the pure information without any redundancy. This paper presents a universal compression system which is able to compress colour images efficiently with a broad range of characteristics from natural images to synthetic images with sparse histograms. The proposed compression scheme uses a novel adaptive parameter selection and a pre-coding stage, which utilizes not only local correlation of the signal values, but also exploits long-distance dependencies. The new approach was tested against several state-of-the-art compressors for a set of seventeen images demonstrating that the average compression efficiency of the proposed scheme is superior.