Sunday, July 16, 2006

Book Fold Correction

Fragile books can't handle being pushed onto glass for photocopying or scanning. These should be photographed as they lie open, without pressure. However, without pressure the pages won't lie flat, so scans look like figure 1. The camera used is sometimes called a 'planetary' camera. They are expensive.

Warped text is fine for reading, but isn't suitable for reprinting or OCR. Some scanning outfits (such as Archive CD Books can correct for book fold warping in software. (I don't know of any software that can do this.)

Google Book Search does this, but often gets it wrong. Google tries to make lines vertical, horizontal, or at a 45° angle. For example, BMC Parthia plate 7 introduces a bizarre 45 degree angle at the bottom right.

Sometimes a few pixels get radically stretched to make a line straight, as BMC Corinth plate 10. That plate is also missing a large rectangle including coins 10 through 13. I don't know if they were removed as part of fold correcting or as part of white space determination. Google also tries to figure out what is text (versus pictures) and replace grey scales with black and white, removing spots in the process. In the case of plate 10 they got it wrong.

No comments: