Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  

FM Forums | Photo Critique | Join Upload & Sell

  

Archive 2012 · Downsizing exercise.
  
 
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #1 · Downsizing exercise.


A reduction from the native resolution of 5616x3744 to a web size of 900x600 results in a linear compression of 6.24 to 1 or 38.9 to 1 in terms of area. Detail is lost and gone forever.

I find some subject matter harder pressed than others. Aspen leaves are a prime example. Aspen leaves have hardly any internal texture to start with but they do have some color gradients as witnessed in my 100% crop.
The minimum processing version was done per this workflow

Note, I donít count this image as a keeper, just a sample for this exercise.

http://ben-egbert-photo.com/?page_id=113

With Topaz photo pop at 70% as shown here.

http://ben-egbert-photo.com/?page_id=344

Topaz Pop messes up color in the clouds and sky so I first mask it and after Topaz. I darken just the sky/clouds on the mask before flattening. (Thanks Karen for mask help)

I believe the version shown here at 100% is sharp enough. With no other adjustments it will print at 16x24 using Qimage default sharpening. It displays well as wallpaper albeit with some compression loses.

I retain the leaf detail of the 100% crop in print and some of it for my 1800 pixel wide wallpaper versions. But at 900 wide they tend to become yellow featureless blobs. Efforts to enhance micro contrast such as large radius small amount USM makes matters even worse. The more I sharpen after downsizing the more blob like they get.

I also note that grass and water and small rocky surfaces get a gritty look if oversharpened.

I have tried lots of web sharpening things in the past including multi step downsize/sharpen and Pixel Genius plug-ins and the various bicubics.

My current downsizing method is to downsize using bicubic (not sharp or smooth) and then apply a very light Topaz Sharpening. The Topaz settings are 0.03 for each category. This seems to cause the least degradation while restoring some sharpness lost in downsizing.

The SOOC version has been converted from raw with everything zeroed. I downsized it with bicubic and no sharpening.

Working on a small image for downsizing is not very practical but posting a full size is not either. If you want a larger size to play with perhaps a 100% crop sooc might work.







Minimum processed full image 900 wide







100% crop of central aspen tree







SOOC downsized to 900 wide




Oct 21, 2012 at 07:18 PM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #2 · Downsizing exercise.


That's mostly why I just work off of my small jpg (3MP) for web stuff as a draft. Then if I really like where I've taken it, I'll work the RAW. OR, I'll work the RAW and not sweat the small web file we post up here cramming my detail together.

Sharpening is a highly "relative" thing, so where you start from, i.e. 21MP to get to around .5MB requires a certain degree of recognition to "engineering constraints".

After all, if you're detail is 4 pixels wide and you squeeze 6 pixels into 1, well kinda tough to save/retain detail within the 4 pixels ... just something that comes with the territory, imo ... with you knowing in your mind that your detail is safe in your real file. With "that much" compression, I mostly accept certain aspects. I'd suggest using an "inset" to show the detail if it is really critical detail that would warrant a 100% view.

I suppose that is in part why I choose smoother when I downsize, then resharpen after sizing (as needed). Once the pixels have been crammed together, it gets tough to separate them again. My thinking is that the smoothing and cramming kinda offset each other before too much damage occurs ... even if it does require a revisit to sharpening after resizing. I never quite understood why the parenthetical suggestion when choosing bicubic versions is the other way around. BTW, I think the "automatic" will go ahead and sharpen it when you downsize anyway (adding to crunchies @ times). It's not really a "neutral" no sharpen/no smoothing option, iirc.

Here's my stab @ the crop ... followed by one sharper, one smoother @ 25% reductions ... noting that I did NOT resharpen the smoother after reduction in the third image ... until I added the fourth image WITH the post reduction sharpening (subjectively applied).

While it might seem like much, I prefer being able to determine how much or how little sharpening to image has, rather than feeling captive to the sizing algorithm, so I typically go smoother / resharpen. That being said, if I'm doing "production work" in LR, then I get one or two dialed in allowing for "sharpen for screen" and follow the rest to a similar level, but that of course is likely a different animal than what we are talking about here.





100%







25% bicubic sharper







25% bicubic smoother







25% bicubic smoother ... resharpen after downsize



Edited on Oct 21, 2012 at 09:52 PM · View previous versions



Oct 21, 2012 at 09:12 PM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #3 · Downsizing exercise.


One other thing about sharpening ...

Sharpening is essentially all about apparent contrast.

This is one reason why I strive to reduce casts first, as they steal from contrast. Also, hue and tonal variance both can provide contrast, so I look to see how much I can use them to drive contrast. These are "in conjunction" with my sharpening, gamma & contrast adjustments, so that I try to drive contrast in "non-sharpening" ways. I also use a variety of threshold levels as well. I rarely sharpen with a threshold of "0" ... I'm typically in the 4-12 range with one round of mid-contrast boost USM and followed by a fine detail round of USM that I take to the edge and then fade.

I find that using a little of each of these things work in concert with a bit less "damage" than sharpening itself.

Of course, like everything ... S&P to taste.



Oct 21, 2012 at 09:46 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #4 · Downsizing exercise.


RustyBug wrote:
That's mostly why I just work off of my small jpg (3MP) for web stuff as a draft. Then if I really like where I've taken it, I'll work the RAW. OR, I'll work the RAW and not sweat the small web file we post up here cramming my detail together.

Sharpening is a highly "relative" thing, so where you start from, i.e. 21MP to get to around .5MB requires a certain degree of recognition to "engineering constraints".

After all, if you're detail is 4 pixels wide and you squeeze 6 pixels into 1, well kinda tough to save/retain detail within
...Show more

Thanks for the examples, I guess working with my 100% crop worked ok for this since its all about the downsizing not how it started.

Working special web versions would totally mess up my work flow, but maybe I could figure that out. I once tried starting with small files during ACR or even in camera, but they are still way too large for the puny web.

What was the sharpening formula for the last version after bicubic smoother?

I see a difference between bi smooth and after it was sharpened, but not between the bi sharp and the smooth+ sharp. First and last.

Thats a huge problem for me, I can't really see fine differences at small sizes. I do see gritty and halo and crunchy and blob, but the trouble is, blobs are almost always there. I am pretty good now at avoiding halo and gritty, but not at retaining detail or even seeing it at these reductions.

I have the Bruce Frasier book on sharpening and had the PK Sharpen program he did before he died and also the Luminance Landscape tutorial for using it by Jeff Schew. So I used to use multi step sharpening and masks etc. But masking and selective sharpen only works for me at full size. Even then I usually use masks to avoid areas for a globally applied thing.

Since I don't currently have any material for artistic critique what I thought I could do is ask some technical questions like this that would be useful to all.








Oct 21, 2012 at 10:36 PM
AuntiPode
Offline
• • • • •
Upload & Sell: On
p.1 #5 · Downsizing exercise.


I'm a bit confused about your sharpening methods, but have you tried readying the image at final size and then making selections, each on a layer, and applying amount and radius appropriate for each selection. If you make the layers smart objects you can switch back and forth from layer to layer and tweak the sharpen sliders to optimally tweak the sharpness of the different selections.


Oct 21, 2012 at 11:11 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #6 · Downsizing exercise.


AuntiPode wrote:
I'm a bit confused about your sharpening methods, but have you tried readying the image at final size and then making selections, each on a layer, and applying amount and radius appropriate for each selection. If you make the layers smart objects you can switch back and forth from layer to layer and tweak the sharpen sliders to optimally tweak the sharpness of the different selections.


My workflow shows my process in detail. Basically I am done sharpening when I am in ACR and before sending to photoshop.

Now I have been sharpening for 10 years and only recently converted to this up front method. I would probably not use that for birds. But for ISO100 landscapes with good focus and sharp lenses, the ACR settings along with Clarity 20 seems sufficient. Topaz Pop adds a bit more and is done with a mask and is usually faded. My 100% crop shows the result, do you think it needs more or less? So long as I have sharp edges and no halos I feel that it is good.

The question here is for downsizing and my process is a work in progress. As I have said, I am not likely to do much selective work after downsizing. The image is too small for my vision and if you blow it up, it is so grainy as to be of little value.

If you have suggestions for that, I would like to hear. The idea of a web only work flow may have some value. Maybe I could do the selective stuff before downsizing and do it differently than for my non web stuff.

I am very happy with me present non web work flow which is driving primarily toward prints. My web process shown here is a work in progress. So far I see no difference between mine and Kents which might just mean poor vision on my part.




Edited on Oct 21, 2012 at 11:51 PM · View previous versions



Oct 21, 2012 at 11:32 PM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #7 · Downsizing exercise.


Yeah, the reductions were kinda small, but I tried to represent a significant reduction even though it isn't nearly as much as going from 21MP to under 600kb

The main thing I see from the bicubic sharper vs. the smoother + sharpening is that the sharper blocks up more going with sharper only.

I think I've got a Frasier and/or Schew book on sharpening as well. The thing to recognize at formulaic sharpening that you might appreciate is the precision / tolerance aspect of things.

Telling that carpenter to shave .003 off of a three foot long board isn't the same as saying take off .003 from an internal piston in a transmission. In one instance, the .003 is meaningless, in the other one it is disastrous. Sharpening a 21MP file is quite different from sharpening a file that has been reduced to less than 1MB. When I sharpen larger files, I'm typically more around 280-420, .7, 4-12, then when I'm sharpening smaller/reduced files, I might be closer to 80-130, .3, 2-6. in the final round.

I should point out that while you mentioned you use ACR for sharpening ... I do nothing in ACR, going to PS with all ACR settings @ 0. Different strokes, for different folks.

Process control, yet subjective evaluation is required ... particularly when you consider that the inherent contrast of lighting varies from specular to diffuse with changing ambient conditions, combined as well with the variability of the subject matter's inherent contrast.







Edited on Oct 21, 2012 at 11:43 PM · View previous versions



Oct 21, 2012 at 11:34 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #8 · Downsizing exercise.


RustyBug wrote:
One other thing about sharpening ...

Sharpening is essentially all about apparent contrast.

This is one reason why I strive to reduce casts first, as they steal from contrast. Also, hue and tonal variance both can provide contrast, so I look to see how much I can use them to drive contrast. These are "in conjunction" with my sharpening, gamma & contrast adjustments, so that I try to drive contrast in "non-sharpening" ways. I also use a variety of threshold levels as well. I rarely sharpen with a threshold of "0" ... I'm typically in the 4-12 range with one round
...Show more

Once I get a non fuzzy edge, I think I am done. Less is more. But I do understand about contrast, This is why you need a bit of side light and something to provide natural contrast like cracks of color variances. I do add some contrast to my images but don't do much for color other than curves and Topaz Pop.

I need to learn more about gamma. I only see gamma as a slider in the exposure layer. I seldom change exposure here, I usually darken with a layer with multiply or using levels.

I used to do all levels of sharpening following Bruce Frasier. But I recently found that the only real sharpening required was capture sharpening if the image was well focused and had good natural contrast. If not,scrap it.

I will ask you the same, did the 100% crop need more sharpening?





Oct 21, 2012 at 11:38 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #9 · Downsizing exercise.


RustyBug wrote:
Yeah, the reductions were kinda small, but I tried to represent a significant reduction even though it isn't nearly as much as going from 21MP to under 600kb

The main thing I see from the bicubic sharper vs. the smoother + sharpening is that the sharper blocks up more going with sharper only.

I think I've got a Frasier and/or Schew book on sharpening as well. The thing to recognize at formulaic sharpening that you might appreciate is the precision / tolerance aspect of things.

Telling that carpenter to shave .003 off of a three foot long board isn't the same as saying
...Show more

I tried a USM step when I was developing my current downsize action. This needs to be an action that downsizes, converts profiles, changes to 8 bit sharpens and saves to a particular folder like 900 wide FM for example. Or at least thats what I want. Someday we will just show our full size prophoto version. Actually, I do upload full size to SmugMug and they do an excellent downsize and sharpen.

I was using Topaz which can oversharpen like crazy with no halo. I got into a trap and eventually ditched that when I found that current ACR was very good and sufficient for non problem images.I almost never shoot anything with OOF areas (on purpose that is).

So when I was working up my downsizer, I tried USM and various bicubic. I avoided Topaz but then I realized it has finer control over small medium and large. I come up with a very fine resharpen that I like.

I can see another approach with bicubic smoother and slightly more aggressive Topaz.

I will certainly try yours too.






Oct 21, 2012 at 11:49 PM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #10 · Downsizing exercise.


ben egbert wrote:
I will ask you the same, did the 100% crop need more sharpening?


Not to sound evasive ... but for me that's a subjective question related to the mood you want to present in your contrast.

Kinda like asking if you should make your technical drawings with a 2B or and 8B pencil of .05 or .03 width. My preference is to use a 4B .03 or a 6B .05. The 4B is a little darker (than the 6B) and the .03 a little thinner, while the 6B is lighter (than the 4B), the .05 is wider. The visual effect is relative similar ... well defined, yet unobtrusively visible.

That's what I like, but others complain that it's too light. Still, it's what I prefer (makes for clean erasures and easy to photocopy invisible or visible with minor copy adjustment).

To answer your question more directly ... it seemed sharp enough, yet low contrast. If I want my viewer to be able to distinguish the details, I believe that I should help them a bit.



Oct 21, 2012 at 11:56 PM
 

Search in Used Dept. 



ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #11 · Downsizing exercise.


I used a 6B very sharp (real drafters pencil not a mechanical type) for layout work. I could print stuff in the middle of the drawing that could not be seen with the naked eye about .010 high lettering. For normal work I used 3b for lines and 2b for lettering. But I converted to cad in 1980.

I have a blueprint from one of my drawings at home that was from a Lockheed bid on the Lem project (Grumman won it).

But you are correct. Sharpening is subjective. I have been all over the map here. Two years ago my work was so over sharpened that I am embarrassed to post at the forum I used to frequent.

I was using halos as a litmus test for when an image was over sharpened. Topaz fooled me because it oversharpens without halos.

When I started posting around 2004 I would always get the "your image is soft". My first response was to send in my lenses for calibration, then by upgrading to more expensive lenses and then when micro adjust came out that with million dollar lenses solved some of the problem.

I finally realized I had been had by a kind of oneupmanship. Slow learner.



Oct 22, 2012 at 12:07 AM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #12 · Downsizing exercise.


+1 @ real pencil ... I've got an 8B that is lethal, will shred paper if you're not careful.

People often mistake contrast and sharpness. My Oly lenses are very sharp, but lower contrast than some of my Nikon lenses. So, my Oly glass needs a bit more sharpening than do my Nikon lenses. But, that doesn't make the Nikon sharper, and in some portions of the image, the Oly is actually sharper, albeit lower contrast.

Acutance, resolving power, contrast, sharpness, etc. are all interrelated toward controlling the amount of difference between pixels and the transition rate (be it slow vs. rapid). The more rapid the change, the more contrast, the sharper it appears. Of course, pure B&W makes for the maximum contrast potential ... and as we use sharpening to change neighboring pixel values farther apart, this is what creates the sharpening perception. Drive them too far apart and you get halos and cruchies.

And ... as you are well aware of, the contrast of the light illuminating your subject plays a significant role in determining the contrast being captured. So ... with the variables of the lens contrast, sensor AA filter blur variation and lighting variance ... sharpening strategies are not universally formulaic, imo.



Oct 22, 2012 at 01:14 AM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #13 · Downsizing exercise.


RustyBug wrote:
+1 @ real pencil ... I've got an 8B that is lethal, will shred paper if you're not careful.

People often mistake contrast and sharpness. My Oly lenses are very sharp, but lower contrast than some of my Nikon lenses. So, my Oly glass needs a bit more sharpening than do my Nikon lenses. But, that doesn't make the Nikon sharper, and in some portions of the image, the Oly is actually sharper, albeit lower contrast.

Acutance, resolving power, contrast, sharpness, etc. are all interrelated toward controlling the amount of difference between pixels and the transition rate (be it slow vs.
...Show more

Very true, but what I attempt to do is reduce the variables so I can use a formula. I do this because I know I can't trust my eyes to do variable changes.

That last is only partly true of course. Another part is because the only audience that cares is other photographers and when you allow them to judge your work, you will get an infinite number of judgments all negative. Or if they are honest like you and some others here, they will say its subjective (as it is) which leaves those who feel like some negative comment are required or at least some change other than what you have done.

But this is a will o the wisp. Like a cat chasing a flashlight beam. I have been doing that for a long time.

So anyway, I use the same few lenses and reject images with poor contrast, exposure or focus. I know what parts of the image don't stand sharpening. I do have several variances to my formulas mostly for dynamic range. But at the saved PSD file sharpening is not one of them.

If you consider it, the only one who actually views the full size PSD is myself and anyone setting by my computer. If it gets printed, Qimage does some print sharpening geared to the print size and paper.

If it is a screensaver, it gets downsized and sharpened like my web formula. If it is for web, it also gets sharpened. So my 100% crop is not normally seen without some additional work.

Fact is, other photographers only see my web versions which are the worst of all versions. I suppose I ought to work harder on them if I want respect.





Oct 22, 2012 at 01:59 AM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #14 · Downsizing exercise.


ben egbert wrote:
Fact is, other photographers only see my web versions which are the worst of all versions. I suppose I ought to work harder on them if I want respect.


I dunno ... I'm kinda okay with "using your imagination" @ how the print file might be better/different than the web file, particularly when it comes to sharpening, be it over or under. Printing is a different animal. Don't get me wrong ... I dig seeing a killer pic on the web, but I recognize that not everything translates ideally for those who are more focused on producing prints.

I guess I can see the "potential" in a pic even if it hasn't been fully finished ... kinda like looking at a negative and envisioning what it can become. I mean, I'm posting some little file when I've got an untouched RAW file "waiting in the wings" to be "handled with care". I don't sweat too much when it is something that I know I can readily attend to when I work the real file.

Old school espouse the value of a great lab to finish your captures. Some are great at both captures and finishing, others are better suited for one or the other. I keep trying to work on both, realizing that my finishing is not something to be underestimated as to what it can do for my better captures.



Oct 22, 2012 at 02:47 AM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #15 · Downsizing exercise.


Hi Kent, I sort of got myself off topic here as I tend to do. I am essentially a technician trying to fit into an art oriented forum. I have been casting around for ways to work with this and thought that asking questions or starting topics that were intended to advance technique might work.

There are some great post processors here. I want to learn more about smart objects, gamma, how you folks manage to nest so many layers, better masking etc.

I tend to be a minimalist at all things, not just photography. But knowing more is never a problem. As my old boss used to say: "it's never a problem if the sausage is too long".

So you have given me a whole bunch of ideas here. I am going to try bicubic smoother with a whole bunch of different sharpening things. I might even cook up a spreadsheet to calculate radius as a function of reduction.

One thing did occur to me however. I get the impression that you start with a more aggressively sharpened image when you downsize. Mine is minimally sharpened and only gets final sharpening when prepared for another stage like print web or screensaver. Sort of in the Bruce Frasier mode. But still bicubic smoother makes sense. It means you are separating downsizing and sharpening.

I also want to blow up one of those aspen leaves to see how sharp they really are. IE how many pixels in the transition at the edge.






Oct 22, 2012 at 02:31 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #16 · Downsizing exercise.


Ok, here is a screen shot of the minimum processed blown up large enough to get the pixel grid. Not sure how enlarged this is but the pixel grid was all I was after.

I tried sharpening this while blown up and the light dark bands using .3 radius show up. It does add some apparent sharpness.

I also tried a downsize to 900 wide using bicubic smoother and various USM starting with Kents 80,.3,4 and moving the slider around. I then did my Topaz 333 on another layer and I was impressed that it was much sharper without adding any artifacts.

I am not sure I could show this at web size so you just have to trust me. I suppose I could play with other Topaz settings, but for now it works good. I am going to change my action from bicubic to bicubic smoother.

Some notes. I think the edge sharpness here has a 3-4 pixel transition if the squares I see are pixels. I also saw that adding very much more USM tended to kill internal tonal values in the leaves.

I feel that this is sharp enough for a master that is still awaiting output.

Edit. Also note, some of the light dark in the transition is probably a result of sharpening applied in ACR. My setting are 50,0.6,7 mask 50









Oct 22, 2012 at 04:11 PM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #17 · Downsizing exercise.


+1 @ technician vs. artist ... I've been on that teeter-totter at times myself.

Spreadsheet at reduction vs. radius, etc. ... I'd be interested to see it. I made some notes a few years back and did a rough assessment. Not quite a continuous calculus curve, but more like counting by tens to put me in a ballpark range, THEN subjective to taste after that. Precision vs. accuracy vs. significant digits kinda thing. Understand it in detail, then apply it only to the level it matters.

Diggin' the pixel grid. Would be interesting to see the pixel grid of the raw file vs. the pixel grid of the reduced file to compare how a given sharpening protocol renders each differently. I already know the answer in concept (which is sufficient for my needs), but my technical curiosity remains a bit ... maybe more of a suggestion for your self comparison efforts.



Oct 22, 2012 at 11:42 PM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #18 · Downsizing exercise.


Hi Rusty.

I downsized the image to 900 and 1800 bi smoother. One for web and one for screensaver. I then applied several sharpening settings. Including USM per 80,.304 as I recall and then Topaz 333,444,555,666. All as separate layers so I could turn them on and off and compare directly.

I can see the difference between no sharpening and USM, and between USM and 333, and if I really squint, between 333 and 666.

This lack of sensitivity told me a spreadsheet would be a waste of time. Sort of like .003 off the end of a log.

I settled on 444 for web and 555 for screensaver. I will make a blow up of the raw. I will take a look at one of the downsized but its going to be gross by the time the grid shows up.

On the tech thing, I never teetor totter, I always knew I was not an artist.

I was not sure how to prepare this so I zoomed up until I saw the grid, (no clue on the magnification). Then I cropped and had a very small image. So I uprezzed bi smooth until it looked right and did a screen capture. If somebody knows a better way I would like to hear.

Here is the raw blow up. Not sure why the grid does not show. It is almost exactly the same size as the first one. I will see what I can do with the downsized version.









Oct 23, 2012 at 12:05 AM
RustyBug
Offline
• • • • • •
Upload & Sell: On
p.1 #19 · Downsizing exercise.


Ya kinda lost me somewhere along the way ... no worries.

Just so long as you know what you're doing / objectives are. I ran my scientific methods to obtain the understanding I wanted/needed for my purposes. I surmise you'll do the same, even if we don't fully follow along. Lord knows people didn't always follow mine either.



Oct 23, 2012 at 12:26 AM
ben egbert
Offline
• • • • •
Upload & Sell: On
p.1 #20 · Downsizing exercise.


Let me see if I can explain. I save the same file downsized bicub smooth at 900 in one case and 1800 in another. Then I applied my various sharpening steps always to the background layer. Each sharpening version was on its own layer so I could shut them down sequentially. This is how I always compare various processes. Side by side does not work for me and sometimes even this does not.

Now here is the result of an 900 wide version downsized bicub smoother then sharpened with Topaz 444. Just blobs here but you asked.

Edit, I might as well show a version at USM 80,.3,4





blow up of downsized with bicub smooth, no sharpening







same but with topaz 444







usm 80, 0.3, 4




Oct 23, 2012 at 12:32 AM





FM Forums | Photo Critique | Join Upload & Sell

    
 

You are not logged in. Login or Register

Username   Password    Retrive password