I previously wrote:

Another acceptable solution is not to permanently save the first generation raw scans, but to preserve the second generation cleaned up versions which have been deskewed (using the right algorithm!), cropped, and for black and white source converted to greyscale (but kept at 600 dpi.) One reason for full-color scanning of even black and white source material is that due to the color characteristics of the background paper, some color channels may be cleaner (essentially "whiter") and could be used for conversion to greyscale. I learned this trick from the DP forum. </smile>

Although quite a few people at DP and PG know about the color channel "trick" to get cleaner greyscale scans of black and white texts, I thought I'd put up an example for those here who have not seen this (warning: need high-speed broadband or a lot of patience): http://www.openreader.org/color-separation/ The file "374.png" is a full-color, 600 dpi page scan (already deskewed and cropped) of page 374 from "My Antonia". The other images are the red, green and blue channels of this image. The paper was really this brown (Willa Cather chose this general color, plus subsequent yellowing/aging has made it worse.) Note how the red channel is significantly better than the others (better contrast between the ink and the paper which should yield a better result when converted to bitonal) with the blue channel the worst. This jives with the general view that the red channel is usually the best. In scanning black and white documents, it is possible to select one of the color channels for greyscale scanning so one need not scan at full-color. Some pretesting with full-color scans might identify the best channel to pick. Or simply scan at full color and in later processing find the best channel to use and batch convert. Anyway, it's amazing how color channel separation can give better results. Jon