User:Jheald/image help

cf: User_talk:Johnbod/22

Moiré pattern in resampled engravings
(eg almost everything from the British Museum prints collection, eg )

A solution would have been to use Fourier methods to notch out everything near the spatial frequency of the new resolution -- see spatial anti-aliasing, Anti-aliasing filter

Practical advice:.

Alternative suggestions eg.

algorithms & comparisons

Image magick page on resizing. Some examples

British Museum collection of prints
(User_talk:Wittylama)

Hi! I hope it's okay bringing this to you. I just thought with your experience inside the BM, it would be much the best thing to try to get in touch and get your best advice first, given that there might be better or worse ways to go about things, and/or sensitivities involved; and of course that you already have the contacts there.

Just as an interesting project, I thought I'd try to put together some gallery pages on Wikipedia Commons on each of some of the various series of engravings made of Cambridge University in the nineteenth century. For example, the publication the Cambridge Almanack included each year a rather nice feature engraving, which (I think) were also issued as sought-after prints in their own right. I thought it would be nice to create a Commons page presenting the whole series, year by year. In the process I've discovered that even some specialist online print shops seem to sometimes only have the vaguest idea when particular prints were done -- for example this image of St John's Kitchen Bridge is suggested as circa 1820, but is in fact from 1803 the earliest in the series I've yet found; here's another copy with a particularly nice web reproduction, being dated c.1825.

What I can find on online web-sites is spotty. Some engraver's series were published in books, some of which have been scanned by the Internet Archive, which can give one source, but only sometimes and only for some prints. (In contrast Google scans tend to be processed, I presume to optimise OCR and/or storage, in a way that tend to either remove the pictures altogether, or at best heavily degrade them).

Of course, somewhere with an outstanding collection of prints is the Department of Prints and Drawings at the British Museum. For example putting in "Cambridge Almanack" and checking 'images only' in the Collection database search brings up almost the complete series, at least between 1803 and 1828. (They seem to be missing this 1804 image of Queens' College, but otherwise between those dates the run seems to be complete).

However, clicking on any of the BM print department results and then "larger image", to give eg this for the Bridge at St John's, it is immediately apparent that the BM images are of far, far, far worse quality that the gallery images of the same size -- the quality of the sky and the particularly beautiful effect of the reflection in the water is utterly destroyed by Moiré patterns introduced by spatial aliasing.

(While I'm at it, one other glitch I've noticed with the website: on the initial search, either from the image page or the corresponding image description page the "next" button work fine; but navigate to a 'related items' page from the image description page -- eg to see "all objects" made by the print-maker S. Sparrow, and then click to go to one of those description pages, expand that image, and the "next" button no longer works; alternatively from the description page click one of the "all objects" links again, and this time none of the links on the search results page work, but instead crash out to the museum's 404 page, with a URL that seems to have an accidental duplication).

But going back to the issue of the Moiré patterns, there is actually a really simple solution, as our page Spatial anti-aliasing discusses (see especially the first section "examples"). The problem is being caused because the BM public images are being reduced from (presumably) a really really high-resolution master scan that the BM has made, but spatial frequencies in that really detailed image are clashing with the periodicity of the low resolution of the final image. This can be cured by blurring the really detailed image with a sinc filter before reducing it. This gets rid of the detail which is causing the problem, and which ought to be invisible in the reduced image.