Wikipedia:Featured picture candidates/Viroconium Cornoviorum excavation

Viroconium Cornoviorum excavation

Reason

High resolution historic photograph from the excavation of the fourth largest Roman settlement in Britain. Contains more material in situ than a comparable modern photo. Restored version of File:Excavation at Uriconium by Francis Bedford.jpg.

Articles this image appears in

Viroconium Cornoviorum

Creator

Francis Bedford

• Support as nominator --Durova³⁰⁸ 16:13, 2 September 2009 (UTC)[reply]
• Comment The words "digitally restored" piss me off... Do everyone a favor and juxtopose the restored with the origional, if it exists, so we can be sure what we viewing is close to legitimate. I won't support it until then.
  talk) 19:41, 2 September 2009 (UTC)[reply
  ]
  • That's what Durova did for years, then people complained and specifically insisted that she merely link. Shoemaker's Holiday ^{Over 203 FCs served} 20:19, 2 September 2009 (UTC)[reply]
    • Actually have pretty much always provided the link to the unrestored version. People who use vulgar language to demand differently aren't worth pleasing. Suggest lifting one's mind from the gutter, reading, and clicking. Durova³⁰⁸ 21:35, 2 September 2009 (UTC)[reply]

• Yes, but you used to include a thumbnail as well, which people got upset about for some reason, and specifically asked you not to, as I recall. Not that it ought to matter, since, you know, if you want to compare a subtle restoration with the original, you shouldn't be using the thumbnails anyway. Shoemaker's Holiday ^{Over 204 FCs served} 09:59, 3 September 2009 (UTC)[reply]

• Seriously, did you bother to click even once before complaining? Other versions of this file are hidden away in the evil-y named "Other versions" section of the image's template. DAMN YOU DUROVA YOU TRICKSTER, HIDING THE ORIGINAL FILE AWAY LIKE THAT! Staxringold _talk^contribs 23:45, 2 September 2009 (UTC)[reply]

• Actually you don't even have to go that far. A link to the original is provided in the reason at the top of the nom, which should always be read before going any further. --jjron (talk) 08:49, 3 September 2009 (UTC)[reply]

• Fwiw, one of the reasons the nominations receive links rather than thumbnails is to encourage close examination. These restorations are performed to hold up at scrutiny of 200% resolution and upwards (faces often get attention at 500% or more). It isn't possible to gauge very much about a restoration from a thumbnail view: thumbnails conceal poor clone stamping and fail to reveal whether dust and scratch removal was performed adequately. In this instance the most delicate part of the work was gleaning more information about the ruins from a very contrasty original. A bit more of the masonry emerges from the shadows, which is interesting because of subsequent alterations to the site. Durova³¹⁰ 16:07, 3 September 2009 (UTC)[reply]

• Support Staxringold _talk^contribs 23:45, 2 September 2009 (UTC)[reply]
• Support GerardM (talk) 16:34, 3 September 2009 (UTC)[reply]
• Comment Not 100% sure about the EV here but one thing is for certiain: this site is well over 500 miles from Caithness! --mikaul^talk 21:56, 3 September 2009 (UTC)[reply]

• • Good catch. The image should be in Wroxeter or Cornovii_(Midlands) but not Cornovii_(Caithness). I removed it from the latter article, but don't have the time to choose a good place in the two former articles. Ksempac (talk) 06:51, 4 September 2009 (UTC)[reply]
• Comment Increasing the saturation confirms what I thought I saw: the chroma noise is much worse than in the original. Since this adds nothing to the image, would it be possible to reduce it?  —SMALLJIM  11:14, 4 September 2009 (UTC)[reply]
  • The original suffered from a small amount of chromatic aberration. That's a scanner problem with limited options to fix in editing. I reduced the saturation by more than 30%; the shadow/highlight adjustment may have produced the effect you saw. It would be possible to get rid of it by editing from only one color channel or by grayscaling the image; either option has drawbacks. Durova³¹⁰ 14:14, 4 September 2009 (UTC)[reply]

There's a degree of CA, certainly - around the tops of the pilae for instance, but there's also an overall increase in chroma noise. The accompanying image shows 400% magnifications of part of the image: left - your version, right - original; top - as displayed, bottom - with saturation increased by 100% three times in gimp. While there's some chroma noise in the original it's well controlled and doesn't stray far from the base colour. However, the processed version seems to shimmer with subdued colour and increasing the contrast shows why: the whole image is covered with a matrix of sharply-defined multicoloured patches. I believe that noise reduction of the chrominance channels should get this back under control. Am I being too picky? I know you do lots of excellent restorations!  —SMALLJIM  16:28, 4 September 2009 (UTC)[reply]

Update: this process seems to work.  —SMALLJIM  16:48, 4 September 2009 (UTC)[reply]

• Taking a try at this now. Staxringold _talk^contribs 20:08, 4 September 2009 (UTC)[reply]

• Comment Alt 1 uploaded with that filter. Staxringold _talk^contribs 20:19, 4 September 2009 (UTC)[reply]

• Yes there's a huge difference – the sky is now green! Really interesting observation but I'm not sure the problem would be so obvious in real-world viewing that it was worth attempting to correct. --mikaul^talk 21:48, 4 September 2009 (UTC)[reply]

• I had rather better success doing that process manually and slightly blurring just the A channel. But it got me thinking. Can we trust the short article
  Photo restoration? It sounds about right to me. This image has clearly acquired tainted colour information during the scanning and subsequent processing. So would it not be preferable to discard the colour channels, which don't appear to hold any spatial information relevant to the original image, and recreate what is thought to have been the original colour by "colorizing"?  —SMALLJIM  00:08, 5 September 2009 (UTC)[reply
  ]

• Agree, what's the point of preserving bogus color information? It's a black and white photograph, so why not just change it to black and white and colorize it to match the tone of the original image? Kaldari (talk) 15:49, 9 September 2009 (UTC)[reply]

Not promoted Seems after all that discussion, there is no quorum. --