Which is sharpened best if you can tell

[arith]

DPP (Digital Photo Professional)



Topaz Detail



I did do a hand held HDR of this but the pixel shift with high contrast exaggerated CA.

[Colin Southern]

Hi Steve,

You've got your black clipping point set a little low which is causing a slight loss of contrast across the image, which is reducing clarity, saturation, and apparent sharpness. Hope you don't mind, but I've reset the black clipping point slightly higher, and applied a 60% @ 1 pixel USM -- have a look at it at 100% and see what you think.

[arith]

Yes that's what I did wrong, it looks a lot better Colin. The first with USM or second with black clipping done first?

[revi]

The last one wins, hands down. That said, I don't know what you wanted to compare, but I'm afraid that the original comparison just went down the drain...
Something else that I noticed, the two images are not quite the same, one is slightly enlarged wrt to the other (slightly smaller crop before resizing?). For me as a scientist, something like that would invalidate any sharpening comparison between software (which you seemed to be after, giving the programs used). Let's not mention what a 2nd round of sharpening on one of the two does to the validity of the test...

[arith]

The last one wins, hands down. That said, I don't know what you wanted to compare, but I'm afraid that the original comparison just went down the drain...
Something else that I noticed, the two images are not quite the same, one is slightly enlarged wrt to the other (slightly smaller crop before resizing?). For me as a scientist, something like that would invalidate any sharpening comparison between software (which you seemed to be after, giving the programs used). Let's not mention what a 2nd round of sharpening on one of the two does to the validity of the test...

Yes they are different because one is sharpened in DPP and the other no sharpening in DPP but later worked in Topaz. The cropping comes last.

[arith]

A bit of contrast applied to Topaz Detail since it is sharper on the large size, cropped to a similar aspect to the others.

[Loose Canon]

You know, Steve?

I think I like your last version best. I also like that style of PP when it is done tastefully as you have done.

[Philip Goh]

I find it hard to tell the first two images apart. I think that's down to the fact that the images have been scaled down for display on the web, which probably demonstrates that for most uses a dedicated sharpening software package is overkill.

[arith]

Thanks Terry, and Philip you are right it is difficult to show a difference in this size, in the bigger size it is possible to see differences that makes Topaz in this case more real surprisingly.

[Dave Humphries]

My apologies in advance for jumping on my hobby horse.

To compare these, everyone one must first reset any browser zoom to 100% - Ctrl and 0 (zero) together does this. I know many, including myself, zoom the browser to get bigger text, but if you leave it zoomed when assessing pictures, you're on a hiding to nothing (wasting your time).

Next up; I'm not sure how many people are just comparing the images as they see them here, downsized by the forums software to 700 x 467 with inevitable softening.

Steve has posted these at 1599 x 1066 px, this means you really need to right click each of the images to open them in their own new browser tab, then click each to view full size, which gets the scroll bars.

An even better alternative to the multiple tabs idea is download each image to your own HDD, then open them all in the PP program of your choice, viewing each at 100%.

If this 100% viewing hasn't been done, I think we are at risk of mis-leading Steve with our replies about sharpening techniques.

Once you have all four images open on separate tabs (or PP windows), you can click between tabs while viewing certain areas of detail and draw some comparisons, here are mine, done from FireFox (hence without the benefit of a histogram, etc.) ...

To my eye, the first two do lack local contrast enhancement, which always detracts from perceived sharpness, especially when viewing the image as a whole. I do see a small difference in sharpening between them, but no more than could be fixed with a little more amount in #1.

Colin's version (#3) gets my vote as the best sharpened.

Steve's last one (#4), at first sight, at 700px, I thought 'that looks better', but when viewed at full size; the buffer beam (red) paintwork looks drastically oversharpened to me, although elsewhere, the effect is less pronounced (and more acceptable).

OK, that's my thoughts on them (and unsolicited guidance on comparing pictures for sharpness).

One parting thought on the picture; unless you particularly wanted to show the uneven-ness of typical railway locomotive bodywork, poor old D182* has been caught in an unflattering light (sun glancing across front) and would have looked nicer had the picture been taken an hour later.

For the anoraks (British term of 'endearment' for trainspotters): D182 is a 'Peak' class, (TOPS class 46) and served as 46045 in later years.

Hope that helps,

[arith]

Cheers Dave;

Of course I meant full size to mean 4545 x 3030 and the buffer does look odd and I sort of gave up trying to get it to look right at 1560 x 1050 or whatever it came out as and also only level 8 in jpeg quality. It is in fact very mottled and looks like it might have been very rusty before heavy painting.

I keep forgetting that an hour can make all the difference and when I get warm I start making all sorts of mistakes; but there isn't a guarantee this will still have been there an hour later since they are being moved around.

Originally I planned a two frame hand held HDR since the EV span was 9.675 according to picturenaught but found the CA (that thing I keep banging on about because of my habit of taking shots in the midday) gets exaggerated by pixel shift. I have in fact done the offending image though it is not very good it took about four hours to recolour.
http://public.fotki.com/Arith/second...co-hf-jpg.html

But I also did the ordinary LDR first at the best exposure and used DPP to add fill light and since I was adding light forgot to clip it, however after seeing Colin's chose to just add more contrast when using Topaz which hasn't got a memory so different Topaz ed images are done similarly, these are not sharpened in DPP since that just leaves a lot of white dots in shadows.

Only the first is sharpened in DPP and since I've seen Colin's I'm wondering could that actually be the best sharpening, but remembering full size means the number on the front fills my 20" screen.

[Philip Goh]

My apologies. Being new to the forums, I wasn't aware that you could click on the images to get a larger size!

Having looked at them, the images I prefer are in the following order: 3, 1, 2, 4.

When I looked at the two original images, I was surprised to find that the DPP version was the one I preferred. Given that the two images I like have been sharpened using bog standard USM, I feel my initial reaction is still valid - a dedicated sharpening tool is overkill.

[arith]

Topaz is normally used for extreme sharpening or special effects. I've noticed after I've successfully removed the porridge like appearance of distant foliage and if I can still read lettering I was no longer concerned with the clean sharp image acutance sharpening gives. But I'm the only one it seems and although a lot on here are much more abstract than I like to be their images are always clean and sharp.

But I paid for Topaz and since it doesn't often introduce halo's thought I would like to try and use it to sharpen in the cleaner sense which involves very little movement from zero.

This is not extreme in Topaz but is an example never to be seen anywhere else:



This is what to me Topaz ed usually means.

[Dave Humphries]

My apologies. Being new to the forums, I wasn't aware that you could click on the images to get a larger size!

No worries Philip, that's partly why I wrote it - not just for you; I mean for people new to the forums. No apology needed.

The earlier version of forum software put the sizes in text under the pictures, so it was far more obvious when clicking would reveal a bigger image, unfortunately, this version doesn't. Also by default (i.e. left click), it opens the image in a Lytebox, which is nice, but (it is a big 'but') it auto sizes to the browser window, so still doesn't give a 100% view for all monitor sizes - only opening in a new tab off the right click can do that (AFAIK).

Cheers,

[PopsPhotos]

This is why I usually upload to tinypic and then use a link.

Pops

[willgoss]

I find the discussion interesting.....It seems to me we are comparing perceived sharpness...
I don't consider Topaz Detail as a sharpening tool but a tool that can increase Detail and sharpness perception....
I would use Topaz Adjust where we have a Sharpness and Detail sliders for simple quick sharpening....

But just my not-so -technical opinion......and a long time Topaz user.....

[Philip Goh]

It seems to me we are comparing perceived sharpness.

That is because all sharpening algorithms increase the perceived sharpness. You can't add detail to an image that wasn't already there to begin with.

[arith]

Yes, and reduce sharpness around an object to increase perceived sharpness. I think the perception of the whole as sharp is more important than individual parts being sharp, except where it is important for those bits to be sharp.

It is all a bit academic as regards the web, since almost everything I do is big and wide, I try to imagine what it would look like six feet wide, and I think resolution sharpness more important there.

So I still don't know, to continue without regard for normal tastes or pull back a bit and do #3 or #4's.

[Klickit]

OK, so I always open images to be compared in seperate browser windows to view and with good results this way.

But why this:

To compare these, everyone one must first reset any browser zoom to 100% - Ctrl and 0 (zero) together does this. I know many, including myself, zoom the browser to get bigger text, but if you leave it zoomed when assessing pictures, you're on a hiding to nothing (wasting your time).

I don't get it. If I CTRL-0 my browser, I end up with Steve's image as a 135mm x 90mm image. Trying to see sharp or un-sharp or pretty much anything at all at that size seems to be tricky. I stare and stare at the different offerings but can see so little detail that I'm quickly zooming in again to fill the screen. Or am I missing something?

[Dave Humphries]

Hi Kit,

But why this:

To compare these, everyone one must first reset any browser zoom to 100% - Ctrl and 0 (zero) together does this. I know many, including myself, zoom the browser to get bigger text, but if you leave it zoomed when assessing pictures, you're on a hiding to nothing (wasting your time).

I don't get it. If I CTRL-0 my browser, I end up with Steve's image as a 135mm x 90mm image. Trying to see sharp or un-sharp or pretty much anything at all at that size seems to be tricky. I stare and stare at the different offerings but can see so little detail that I'm quickly zooming in again to fill the screen. Or am I missing something?

If you did the viewing on separate tabs bit right - and clicked the image to overflow it (being 1599 x 1066 px, unless you have a huge screen, it should give scroll bars) I don't see how it is that small.

If you are missing something, it is because I explained it badly, or you're using a browser that operates differently from FireFox, or both

Tell me more ... (about your screen, browser, etc.)

If you use browser zoom (i.e. by not doing Ctrl 0), it WILL look soft and in these images it really isn't necessary, given their large size. Using browser zoom on an image, especially one that is big enough, is akin to paying for an expensive lens, then always shooting through the car window glass and wondering it isn't sharp. OK slightly dodgy analogy, but it is that fundamental.