Cool Tools: Retro Reveal

Submitted by: Jennie Levine Knies

The UMD Libraries’ Lovely Cassandra Schmidt brought this tool to my attention. Retro Reveal. Hosted at the University of Utah’s J. Willard Marriott Library, and utilizing the ImageJ technology developed by the National Institutes of Health for scientific image processing, Retro Reveal helps users uncover hidden text in their documents.

For example, hard-to-read postage stamps.

A useful tool for uncovering dates on postmarks, faded pencil marks, and other mysterious historical ghosts.

I conducted a recent test to see how this might work with one of our out-of-focus newspaper scans that we created as part of our Library of Congress/NEH-funded National Digital Newspaper Project grant.  There were a few kinks to work out. First, the tool only accepts images that are under RGB, 20MB and less than 6 megapixels.  Our newspaper project images are grayscale, often much larger than 20MB, and digitized at a high resolution. I took one of our images, cropped it, changed the color profile to RGB using Adobe Photoshop, and lowered the ppi from 400 to 200.  This gave me a 17.5MB image of a page from Der Deutsche Correspondent.  It is difficult to evaluate how useful this tool is for this page.  A more accurate test might have involved simply uploading several small, blurry subsections.

Cool Tools: Bodleian Ballads image-match demonstrator

On more than one occasion, a researcher (most notably, my husband, who is studying 19th-century printing), has asked me if I knew of any method for quickly identifying or comparing images. The Bodleian Ballads Image Match seems to be just the type of tool for which I was searching. The tool was “experimentally developed under a John Fell Foundation grant awarded to Dr Giles Bergel (Oxford Faculty of English) and Dr. Richard Ovenden (Bodleian Library), with the assistance of Dr. Alexandra Franklin (Bodleian Library). High-quality (24-bit, 600 dpi) TIFF images of a sample of some of the library’s 17th-Century ballads (when the form was most copiously illustrated) were made in-house, courtesy of Bodleian Imaging Services, and indexed by the Visual Geometry Group.” “Instance-level recognition,” is the technical description  of this process, and equivalent to some degree of optical character recognition (OCR) for text.  I can see all sorts of potential for utilizing this technology in research, such as identifying common images in a project such as Chronicling America, or pulling together similar William Morris woodcuts.  Watch a video demo describing how the Bodleian Library is now integrating image-search into a redeveloped Bodleian Broadside Ballads database, launched in 1997, as part of its JISC-funded Integrating Broadside Ballads Archive Project, a collaboration with the English Broadside Ballads Archive at the University of California at Santa Barbara and the Vaughan Williams Memorial Library at the English Folk Dance and Song Society. How might we use this tool at the University of Maryland Libraries?