Joint entropy and multiband prediction for lossless image compression

The transmission and storage of a digital image requires considerable expenditure in terms of bandwidth and/or memory resources. Digital image compression is concerned with minimising the number of bits used in representing an image, and is therefore a means of reducing image data for economical storage or transmission. Image compression techniques can be grouped into two broad classes: (a) lossless techniques, and (b) lossy techniques. Some applications require lossless compression of the image data, e.g., remote sensing, the main application area of the work described in this paper. For multiband satellite images redundancy exists in three forms, namely, statistical redundancy, correlation redundancy, and interband redundancy. The reduction of statistical redundancy and correlation redundancy within an image has been extensively investigated by many researchers. The paper explains that correlation redundancy was first exploited within each band of a multiband image. A new stage exploiting interband redundancy was then introduced. The final stage was the reduction of statistical redundancy. Interband redundancy is quantitatively analysed, and two schemes are developed to exploit it