Quantization Range Conversion Using  Nonincreasing Filters

doi:10.1117/PM29.ch6

Quantization Range Conversion Using Nonincreasing Filters

Book: Enhancement and Restoration of Digital Documents: Statistical Design of Nonlinear Algorithms

Author(s): Robert P. Loce, Edward R. Dougherty

Published: 1997

https://doi.org/10.1117/PM29.ch6

Author Affiliations +

Abstract

Digital documents are typically created, stored, transmitted, and printed at 1 bit/pixel. Although the image class is binary in nature, contour information present in the ideal form of the document may not be well represented in this sampled form. (See Section 1.3 for a discussion of the relevant imaging issues.) In this chapter we describe the use of nonincreasing filters for estimating a more finely quantized image from a binary image. One class of filters described here is an estimator of the gray-scale conditional expectation of area samples quantized to, typically, 2, 3, or 4 bits/pixel. Another class takes into account the image writing mechanism and uses additional quantization levels to identify partial area coverage of a pixel formed by a binary marking process. We show how these filters can play an image enhancement role in an electronic or digital display printing system. These filters estimate the digital image quantized to M bits/pixel prior to printing. The effect on the overall system output is a smoother rendering of contours.