Efficient facsimile data reduction by thinning process

Abstract

Picture preprocessing and encoding methods that can greatly reduce the facsimile picture signal information are presented here. Preprocessing consists of smoothing and thinning lines before encoding in a transmitter. Lines are thickened by postprocessing after decoding in a receiver. Because the post-processed picture quality is lower than that for the original picture, it is important to prove whether entropy reduction by thinning is still great. To do so requires considering picture quality degradation caused by preprocessing. Proof of high entropy reduction involves a comparison of the entropy for a preprocessed picture with that for a reduced-size picture. The picture quality of the latter when enlarged back to its original size is comparable with that for the post-processed picture. This comparison reveals that entropy reduction by preprocessing is 5 percent for printed Japanese and 19 to 24 percent for handwritten English and Japanese, as well as for typewritten English. An encoding scheme called predictive MH coding, which takes advantage of the feature wherein line width in a preprocessed picture is almost a single picture element, is applied to encode preprocessed pictures. As already predicted by entropy measurement, it was found that the compression ratio for the proposed method is 1.04 times as high as that for the CCITT-standard MR coding for printed Japanese. Furthermore, it is from 1.3 to 1.4 times as high as that for the CCITT-standard MR coding for handwritten English and Japanese, as well as for typewritten English.