Decision feedback equalisation for document input

TL;DR: It will be shown that DFE can be used to obtain a binary image with the same quality of a surface which is 2 to 3 times as large as when straightforward thresholding is used.

Abstract: Obtaining digital images from binary scenes (for example text on paper) and subsequent thresholding does not necessarily lead to high quality binary images. In the imaging process. the binary scene is low-pass filtered and tiny details are therefore smeared. This low-pass filtering is necessary for the subsequent sampling to get a discrete image. In this paper, we bomw from the tho ry of data transmission, where a similar smearing process occurs: inter symbol interference (ISI) [Qu85]. We have tried to use the theory of decision feedback equalisation @FE) in order to improve the quality of detail of binary images from binary scenes. It will be shown that DFE can be used to obtain a binary image with the same quality of a surface which is 2 to 3 times as large as when straightforward thresholding is used. The method will be compared with dynamic thresholding [Be861 and iterative constrained deblurring [BL90]. Another application of DFE in images is in image enhancement