Multilevel graphics representation using block truncation coding

In this paper a modified version of a noninformation preserving coding technique known as block truncation coding (BTC)is used to code multilevel graphics. The method is based on-local binarization of nonoverlapping blocks in the graphics image such that the sample mean and sample variance in each block is preserved. BTC is easy to implement, does not require multiple passes through the data, does not degrade when compressing noisy images, and can be made relatively insensitive to channel errors. Examples of the coding technique are given and compared to other compression methods. This technique shows promise for coding 32-level (or 32-color) graphic images at data rates of about 1 bit/pel.