Icones Avium Dev Diary #2: Source material

Digital restoration involves cleaning up all the foxing, spotting, stains, rips and tears, removing colour casts and balancing the levels to maintain the natural depth of the original image.

I’m keen that my digital editions of old natural history books should have high quality images. There’s no point doing this if the images aren’t up to scratch, and restoring digital images requires a high quality source. The best option is of course to scan a physical copy under controlled conditions. But for rare, old manuscripts that’s usually costly and impractical.

In the case of Icones Avium, one was sold in 2017 for $18,750USD, relatively cheap for a John Gould original. But such a cost is beyond an independent enthusiast like me, so instead I turned to the public domain to find a secondary source.

Obtaining a good quality source

I found three online versions of Icones Avium.

• Icones Avium supplied by Biodiversity Heritage library (BHL)
• Icones Avium supplied by National Sporting Library & Museum (NSLM)
• Icones Avium supplied by Kansas University Digital Collection: Part 1 and Part 2 (KU)

I’ve assessed these sources for their image quality. While the quality is good overall, they all suffer from various problems that impede a high quality restoration. Issues include high JPEG compression, foxing and spotting, tight crops, and scans taken from books that were not flat, distorting the shape of the birds.

A comparison of the three source scans, using plate #3 Ianthocincla Phoenicea, the Scarlet-faced Liocichla. From left to right the sources are BHL, NSLM and KU.

Biodiversity Heritage library

The images from the Biodiversity Heritage Library are undistorted but are compressed in JP2 (JPG2000) format which loses some fine detail. This could be offset by the fact these images have the highest resolution. The colour cast of these images is on the yellow side, but this is easy to remove.

National Sporting Library & Museum

The images from the National Sporting Library & Museum fare less well. These are in JPG format, which produces worse compression artefacts than JP2 (it’s very noticeable at high zoom). Some of the images are very tightly cropped, in one case part of a tail is missing. And they are all very dark. I feel that these are the poorest quality scans out of the three, so I’ve eliminated these from the decision.

Kansas University Digital Collection

The images from Kansas University have the best scan quality, with uncompressed TIFFs scanned at 600dpi. However the book isn’t flat and many of the images are distorted. They also suffer from a blue colour cast (although this should be easy to remove).

Should I therefore use the high quality Kansas University images and accept the distortion, or use the BHL images which were not distorted but had compression artefacts? I can’t make a decision based on these factors alone, so I’ve also assessed the quality of the hand colouring. And that’s where it’s easy to choose a clear winner: the colouring of the BHL source is far inferior, lacking detail and somewhat washed out. The Kansas University images have much more vibrant colours, are more detailed and seem more precise overall.

I’d love to use the Kansas University images. They’re evenly lit and the paper is neutral and very clean. These scans would provide the best outcome during the restoration process. But the curvature of the pages has created inaccurate images which is far from ideal. It would be possible to reverse this using a mesh distortion in Affinity Photo, but the tool is imprecise and I don’t fancy my chances of a great result.

I’m deferring the decision for now: I will restore two or three images from both sources and see how they look once I’ve got a bit further into development of the prototype.

The tail of the pennant-winged nightjar is so long that it doesn’t fit on an imperial folio sized page (15″ x 22″)

An interesting detail for this project is that Icones Avium includes one plate on a folding page. The pennant-winged nightjar has a remarkably long tail, and depicting this at life size on an imperial folio page was impossible. An extra large page was used to provide the space for the tail, and folded over when the book was closed. This oversized image will cause problems with image presentation consistency in the digital version, but that’s part of the challenge.

In the next dev diary, I’ll explore how to structure the content of the project.