Scaling of halftoned images

Abstract

The goal of this paper is to describe technique for inverse screening halftoned images, created for an electrophotographic printer, and scaling the contone result. The objective is to achieve high quality results on another printer, for example an ink-jet printer. The approach we take is in the spatial domain. However, screen characteristics are determined in the frequency domain. Deducing screen attributes from the samples is possible and even quite robust. Test results yielded very good estimates. We found that for a 96 LPI screen at 51 degrees, where the spatial vectors for the screen are v1=4, 5 (horizontal, vertical) and v2=-4, 5 - the estimated parameters are in agreement with the data within 1.5%. This means that we can identify the closest screen and use predefined data for our solution. For an unknown screen we can produce the data on the fly. Next step is to convolve the image with a flat filter (of only ones and zeroes) at the correct shape and size. This yields very nice results having excellent contone reconstruction which are better than a Gaussian low pass filter. We'll show how this reconstruction procedure is connected to the actual halftone properties, thus fitting our intuition as well as human visual system (hvs) properties. Next we scale the image to avoid moiré patterns at the contone level.