Talk:Black-winged Stilt
From Wikipedia, the free encyclopedia
Image:Black-necked Stilt-252.jpg
Manual thumbnail.
I have a problem with the new thumbnail code. It just doesn't produce a quality result. Compare the two images. The difference speaks for itself. Tannin 12:30, 5 Feb 2004 (UTC) Errm... click on enlarge: the large image already looks fuzzy to me... Lupo 12:39, 5 Feb 2004 (UTC) Is that better? I sharpened the large image marginally. The small image (software generated) should also improve now (seeing as it has a sharper original to work from) but it still won't be nearly as good as an image generated by a system with a human being in the loop. (You may have to clear your cache to see the differences.) Tannin Ah... "it still won't be nearly as good as an image generated by a system with a human being in the loop" — you are aware of that. Apparently I had misunderstood; I thought your problem was that a human could outperform the software on a visual task... I don't think the SW is to blame in this case: its task is to produce thumbnails, and of course those will be fuzzy if the original is fuzzy. It's not the task of the SW to automatically sharpen an image. So, where's the problem? :-) Lupo 12:52, 5 Feb 2004 (UTC) Quite so: a re-sized image nearly always needs sharpening and the software can't sharpen the image, as the required amount of sharpening varies from one image to the next. It needs the eye of a human being to judge the correct amount, and also to apply any other corrections needed. For example, in the stilt thumbnail at right, I also upped the gamma just a fraction, to give the image a visual lift. (Obviously, a serious PhotoShop jockey would do a good deal more than that, but that takes a great deal of time and/or skill - what I did took about five mouse clicks.) Also, a human can select a good reduction ratio, as even-number shrinks produce the sharpest, clearest thumbnails (i.e., a 50% or 33.3% shrink looks better than a 54% or 29% shrink) and it's often possible to find an even % shrink number that is close enough to the desired size for practical purposes and yet still retains as much quality as possible.) That's my point: I don't like having the software do the shrinking. It was a great idea, and doubtless some very clever coding) but the images it produces can't compete with the images you and I (as humans) can produce. We are better off not using it. Make sense? Tannin 13:44, 5 Feb 2004 (UTC)
Yes, that makes sense, but I still like it. The software seems to do an adequate job on most images. Maybe it's not perfect for photographs, but other images (diagrams, flags, and so on) should be less problematic. The above is also an extreme case, because it started off with a fuzzy big image. With reasonably sharp originals, I have so far not come across a software thumbnail I found unacceptable—but hey, maybe my standards are just too low :-) About the resize ratio: you do have control over that. Just specify the width in pixels as a reasonable fraction of the original image's size. For instance, the big image is 600 * 624 px. I'd try scaling this down to 300 or maybe 200px, but probably not to 250 or 252px (41.67% or 42%):
Using an integer shrink percentage (252px = 42%; on the right) already gives a slightly better result than the fractional 250px (41.67%) on the left. (Although the shrink factor is fractional then: with 250px, the factor is 2.4, with 252px, it's 2.3809... so what is important here? Factors or percentages?)
When it matters, I play around with the thumbnail sizes until I get a reasonable result. The auto-generated thumbnails for this picture for instance look acceptable to me at 300px (50%) or 270px (45%). And when it really matters, I'd just do what you did and produce a manual thumbnail. Lupo 08:42, 6 Feb 2004 (UTC)
[edit] Range
My book gives the southern breeding limit of Black-necked Stilt as southern Chile, S Argentina and the West Indies. Is there a split I don't know about, or is the range just wrong? If the latter, the "where to see bit becomes very parochial". jimfbleak 06:35, 24 August 2006 (UTC)