On turbo-compression and modeling of black and white images

This paper considers lossless compression of black and white images using recently developed distributed turbo compression techniques. An empirical study of image pixel statistics is initially conducted in order to apply turbo compression algorithms to distributed compression of image rows. In particular, it is noted that while the entropy of the pixels throughout the image does not vary much, the distribution and conditional distributions of pixel values are nonstationary. Corresponding compression bounds are noted, appropriate coding schemes based on turbo-codes are developed and simulation results are reported.