Yasuo Kurosu

Evaluation of effective data reduction methods for MMR coding

To obtain higher speed and higher quality of image communications economically, a standard coding method with a higher efficiency is required. This paper proposes three methods to eliminate horizontal-mode codes (which are mainly contained in noise components) from each coding process. The idea is based on the fact that MMR coding can be highly compressed by rising 2-D correlations. The three proposed methods are compared in respect of compression rate, processing speed, and image quality by using document images. The results show that the run-length code reduction method is best, the compression rate being higher than conventional methods by 19.6% (maximum) and 12.7% (mean), the processing speed being 2803 times that of conventional binary-image processing, and the image quality being high due to noise elimination.