Detection of Monochrome Pages in Color Document Scans

Abstract

In this paper, we present an algorithm for detecting monochrome pages in a color copy job on a multi function printer (MFP) with a contact image sensor (CIS) based scanner. Once detected, a monochrome page can be processed as such, which can improve image quality, print speed, and save color-printing supplies. The presented algorithm processes the RGB color data captured with the CIS scan bar for a given scan band and keeps track of color information across all scan bands moving down the page. The colorfulness of every pixel in the scan bands is derived from the Cb and Cr channels after color conversion from the RGB to the YCbCr space. A pixel is classified as color if its colorfulness value is greater than a predetermined device and media specific threshold. This threshold is found by modeling the cumulative colorfulness histogram of a number of scanned test documents using a mixture of two Gaussian distributions and the Expectation Maximization (EM) algorithm. For every scan band, the highest concentration of color pixels is saved, and later used to classify the page content as either color or monochrome.