(The) performance of a multilevel runlength coding for multilevel image

Abstract

The performance of an extended conditional runlength code(CRC) is investigated for multilevel images. The new coding method named as "Multilevel conditional runlength (M/R) coding" is basically an extension of CRC using both runlength code and Huffman code based on the statistics of image states. In M/R coding, after the fundamental states of Markov model have been classified into several groups with different probability distribution, runlength coding or Huffman coding is appropriately used in each group to achieve a pel entropy which approaches the Markov bound. In this paper, the effects of various parameters are investigated to minimize the error propagation within the specified image quality for 6-bit images. In comparison, it is found that M/R coding is superior to Bit plane coding, and is comparable with Discrete cosine transform.