USM second opinion

On Sat, 4 Nov 2006 10:39:40 -0500, "Phil M" found these unused words floating about:

To my eyes, the "heavy" version looks better when the image is viewed entirely. By contrast, the "medium" setting looks with less artifacts at 100% but a bit soft when viewed entirely. I would like to get a second opinion, which is better? (pleasing without oversharpening) The medium or heavy setting? Thanks.

Are you talking about "100% viewed entirely" as viewed in a print -=OR=- as resampled by your monitor at a lower ‘viewing percentage’?

If the latter, you’re losing resolution by combining pixels to ‘shrink’ the image to fit. NOT a method for comparison !!!

Are you talking about "100% viewed entirely"

By "100%" I mean actual pixels on the monitor. No resampling. Obviously, at 100% magnification I see only part of the photo.
By "viewing entirely", I mean shrink to fit the entire image on the monitor which is roughly 35% magnification.

When comparing both images at shrink to fit on the monitor, the medium is soft;
When comparing both images at shrink to fit on the monitor, the heavy looks good.
When comparing both images at 100% on the monitor, the heavy has artifacts; When comparing both images at 100% on the monitor, the medium looks good.

The final purpose I plan to do is not to print; I plan to project them using a computer projector AND create a slideshow for monitor viewing. Thefore, I’m looking for a second opinion on USM settings. Thanks.

I read many opinions on the subject of USM.

You should also read about posting images for the web.
This images is much to large
Do shift/cntrl/alt/s – image size 800X600 or smaller
And the last image is over sharpened.
But, this is only according to my pref – I sharpen very moderate

I am flying (owning) a Cherokee 180 –
this may be a Cessna 172’s engine

Dave
http://home.intekom.com/davesplace/durban.html

Do you realize that sharpening should only be applied just prior to printing and tailored to the specific requirements of that print? There are many, many ways to sharpen images and the degree to which images should be sharpened is entirely a subjective judgement.
If you use CS2 it is difficult to find that alternate sharpening methods are preferable for general use to the smart sharpen tool.
This is a link to a tutorial on the smart sharpen tool created by someone with far more expertise than I in Photoshop
http://www.radiantvista.com/video_tutorials/page3/ which you may or may not find useful.
Also you might want to think about processing your image a bit further before even considering printing/sharpening: remove dust, optimize contrast, consider noise reduction, etc.

You should also read about posting images for the web.

Actually not for this case. Sharpening is resolution dependant. If I lower the rez, then different settings would be required.

And the last image is over sharpened.

Thanks.

I am flying (owning) a Cherokee 180 – this may be a Cessna 172’s engine

PS. For the record, this is a Piper Cherokee N718FL which turns out to be a Lycoming 320 series engine…. Although I tried Photoshopping my 1998 C172P panel to add Garmin G1000 and make my friends jealous, but it didn’t work. On my next attempt, I’ll follow your advice to issue a low rez version (damn, that’s why they cought me with my high rez imagery…) Cheers. 🙂

On Sat, 4 Nov 2006 12:33:58 -0500, "Phil M" found these unused words floating about:

Are you talking about "100% viewed entirely"

By "100%" I mean actual pixels on the monitor. No resampling. Obviously, at 100% magnification I see only part of the photo.
By "viewing entirely", I mean shrink to fit the entire image on the monitor which is roughly 35% magnification.

When comparing both images at shrink to fit on the monitor, the medium is soft;
When comparing both images at shrink to fit on the monitor, the heavy looks good.
When comparing both images at 100% on the monitor, the heavy has artifacts; When comparing both images at 100% on the monitor, the medium looks good.
The final purpose I plan to do is not to print; I plan to project them using a computer projector AND create a slideshow for monitor viewing. Thefore, I’m looking for a second opinion on USM settings. Thanks.

FWIW: I’d probably go with the ‘medium’ for use with the projector -=IF=- its resolution matches the image resolution. (Most projector’s ‘fit’ algorithm are useless).

I’d shring the unsharpened for the ‘slide show’ (monitor) then decide about sharpening.

As before, you can -=only=- compare at full resolution – the scaling is a quick and ‘dirty’ -=temporary=- function – NOT for judging!

Also you might want to think about processing your image a bit further before even considering printing/sharpening: remove dust, optimize contrast, consider noise reduction, etc.

Thanks for the suggestion. The image prior to sharpening (no-sharp.png) has been already been optimized: Infrared ICE4 for dust/scratch removal, Kodak ROC white balance, Curves/levels and Noise Ninja Noise Reduction.

On Sat, 4 Nov 2006 17:39:24 -0500, "Phil M" found these unused words floating about:

Also you might want to think about processing your image a bit further before even considering printing/sharpening: remove dust, optimize contrast, consider noise reduction, etc.

Thanks for the suggestion. The image prior to sharpening (no-sharp.png) has been already been optimized: Infrared ICE4 for dust/scratch removal, Kodak ROC white balance, Curves/levels and Noise Ninja Noise Reduction.

Why are you saving in a lossy format -=IF=- you intend to do further work on it?

Why are you saving in a lossy format -=IF=- you intend to do further work on
it?

png is not a lossy format.

I’m really just re-iterating bmoag..

Phil M wrote:

I read many opinions on the subject of USM. Most Nikonias recommend a setting of 150/0.9/0, others advocate 300/1/4/1, others swear by 20/60/4 at scanning then 500/0.5/0 in PS, while the book "Scanning & Halftones" suggests 200/1/3 as a starting point.

Did you read *all the surrounding text* on those sharpening recommendations? Are they all for a 100% display image? (one of them might be..(O;) A 300 ppi print? A 200 ppi print? A 6"x4" print? A 24" x 36" print? Film scans or dslr images? To be viewed at what distance?

Because the sharpening for all of these will be *different*, as bmoag says. I work with film scans too, but I would never sharpen the scan at full scan res, especially while it still displayed so much grain/aliasing. At the size you have displayed it, that image would print at about 5 feet wide! Is that realistic for the image? I *always* downsample my filmscans to at least half (preferably one third) of their original size, before even thinking about the sharpening level I might apply. That way the aliasing (and other optical issues) are smoothed/averaged/reduced, and sharpening can effectively be applied to the ‘real’ image detail, rather than all the ‘noise’. You will then get a far better feel for the settings – it seemed like black magic to me at first, but I’m getting it now… By the way, some of those settings you list are almost certainly designed for specialised sharpening, eg using LAB mode layers, found edges, multiple sharpening runs, etc..

I tried all and due to the graininess
and particularities of this duplicating film – none of the above settings work.

And I don’t think anything will, until you properly downsample and clean up the image. Sorry!

To my eyes, the "heavy" version looks better when the image is viewed entirely. By contrast, the "medium" setting looks with less artifacts at 100% but a bit soft when viewed entirely.

Another issue here is that when you zoom the image, whether by browser or image processing program or whatever, you are at the mercy of their zooming algorithm, which is almost certainly not going to give you results comparable to doing it properly. Investigate and compare the results you get from different downsampling routines in Photoshop, and compare them to the results from just zooming..!

When the image is very messy, like yours (sorry), I would use heavier denoising first, maybe even a slight blur as well, then a simple downsample to say one-third size. Then try sharpening… When the image is pretty good – eg from a dslr, I use Lanczos downsampling (try Irfanview), and then very lightly sharpen. (Lanczos seems to be the best at holding detail without introducing artefacts – like jaggies along diagonals.)

Sharpening is always the very last thing I do, just before I print or post, and only when the image is at *exactly* the size I want, for the resolution I wish to print or display it..

My sharpening settings would probably range between 50%-300% (usually around 100%, and higher for printing), 0.3 to 3 pixels (0.3-1 for display, 1-3 for high quality printing), 0-8 threshold (higher numbers for noisier images). Display settings are usually very different to those for best prints, and then they can also vary a lot depending on image content.

Finally, the better you get at the entire process, the less you need/want. (O;

At the size you have displayed it, that image would print at about 5 feet wide! Is that realistic for the image?

As a matter of fact, in this case, yes. Because the image(s) are for projection. But the digital projector is only capable of 1280×1024. Very true. Thanks. I downsampled the image to this size and I settled for medium settings. The nasty grain, alias and what have you disappeared.

Finally, the better you get at the entire process, the less you need/want.

I recall reading somewhere: Sit in on a 5 seminars, or read 5 articles on sharpening, and you’ll get 10 different opinions. But everybody seems to agree that Sharpening is like Garlic in Cooking, everyone has their own taste, but too much or too little can spoil a dish.

wrote in message
SNIP

I *always* downsample my filmscans to at least half (preferably one third) of their original size, before even thinking about the sharpening
level I might apply.

"Before even thinking about", indeed !!!

Care should already be taken with regards to the down-sampling method used. Unfortunately Photoshop’s methods are not the best 🙁 . As a minimum, *not* following Adobe’s recommendations, one should use ‘plain’ bicubic resampling:
< http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1 .htm> as a practical scan to smaller output evaluation.

The results above are predicted by this more theoretical analysis: < http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sam ple.htm> ..

That way the aliasing (and other optical issues) are
smoothed/averaged/reduced, and sharpening can effectively be applied to the ‘real’ image detail, rather than all the ‘noise’.

Agreed for down-sampling, one should avoid *adding* more artifacts due to down-sampling, but also don’t use Photoshop for any substantial up-sampling. There are better solutions for that task, and since upsampling usually is used for best possible print output quality I suggest using ‘Qimage’ for that part of the workflow.

—
Bart

On Sat, 11 Nov 2006 04:21:22 +0100, "Bart van der Wolf" wrote:

wrote in message
SNIP

I *always* downsample my filmscans to at least half (preferably one third) of their original size, before even thinking about the sharpening
level I might apply.

"Before even thinking about", indeed !!!

Care should already be taken with regards to the down-sampling method used. Unfortunately Photoshop’s methods are not the best 🙁 . As a minimum, *not* following Adobe’s recommendations, one should use ‘plain’ bicubic resampling:
< http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1 .htm> as a practical scan to smaller output evaluation.

The results above are predicted by this more theoretical analysis: < http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sam ple.htm> .

That way the aliasing (and other optical issues) are
smoothed/averaged/reduced, and sharpening can effectively be applied to the ‘real’ image detail, rather than all the ‘noise’.

Agreed for down-sampling, one should avoid *adding* more artifacts due to down-sampling, but also don’t use Photoshop for any substantial up-sampling. There are better solutions for that task, and since upsampling usually is used for best possible print output quality I suggest using ‘Qimage’ for that part of the workflow.

…you bloody Europeans & Yanks can’t count.
How on earth can you keep on calling this thread the ‘USM second opinion’ if the recent one is about the 139th opinion?

Dave
with a calculator on his desk

Bart van der Wolf wrote:

Care should already be taken with regards to the down-sampling method used. Unfortunately Photoshop’s methods are not the best 🙁 .

Up until about 12 months ago, I thought there wasn’t much skill required to downsample, or at least nothing like that required for upsampling. I am a little wiser now!

As a minimum, *not* following Adobe’s recommendations, one should use ‘plain’ bicubic resampling:
< http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1 .htm> as a practical scan to smaller output evaluation.

This is a *very* useful page – and it made me revisit my methods and take a much closer look at Lanczos for downsampling, which I now use very frequently.

The results above are predicted by this more theoretical analysis: < http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sam ple.htm>

When I first saw that page, I found it hard to believe that many artefacts would be introduced. Thank heavens that sort of detail isn’t found in most images! But it helps explain why the artefacts appear and which methods are best employed to avoid them.

Agreed for down-sampling, one should avoid *adding* more artifacts due to down-sampling, but also don’t use Photoshop for any substantial up-sampling. There are better solutions for that task, and since upsampling usually is used for best possible print output quality I suggest using ‘Qimage’ for that part of the workflow.

Agreed. Or Lanczos, again, I find is a good (and free) compromise, when combined with small-step increases.

Thanks for sharing all your efforts on this topic, Bart.

Lanczos, again, I find is a good (and free) compromise,
when combined with small-step increases.

Can anyone comment how would the Lanczos filtering method vs. Fractals Print Pro scaling algorithm (Photoshop CS plug-in) compares?
(image samples and free trial):

http://www.ononesoftware.com/detail.php?prodLine_id=7

Fractals are believed by many to be capable to upsample images better than any other method. However, for down sampling, nobody has an answer. But, if the upsample algorithm is the best, what would suggest that its downsampling method is worst?

However, for down sampling, nobody has an answer.

I just got my answer here:

http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1 .htm and here
http://photo.net/bboard/q-and-a-fetch-msg?msg_id=00E7py

The best downsampling algorithms are found in ImageMagick with the Laczmos filter. It seems that this freeware blows everything commercial out there… After fiddling with it for 3 hours to learn its basics, I did a test. Photoshop CS2 Bicubic vs ImageMagic Lanczos: The difference is obvious. Now, I just have to learn how to do batch jobs.

"Phil M" wrote in message

However, for down sampling, nobody has an answer.

I just got my answer here:

http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1 .htm and here
http://photo.net/bboard/q-and-a-fetch-msg?msg_id=00E7py

The best downsampling algorithms are found in ImageMagick with the Laczmos filter. It seems that this freeware blows everything commercial out there… After fiddling with it for 3 hours to learn its basics, I did a test. Photoshop CS2 Bicubic vs ImageMagic Lanczos: The difference is obvious. Now, I just have to learn how to do batch jobs.

If you are running it from Win XP, you could start with something like:
For %I in (*.tif) Do convert %I -filter lanczos -resize 100×100 %~nI.jpg
and adapt for the particular batch job at hand if you want to do something
else than convert all Tiffs to JPEGs in the same current directory and fit
in a 100×100 pixel thumbnail while maintaining aspect ratio. Prepending
an output path to the output filename would be a logical thing to do.

—
Bart