Photoshop: Tones not Uniform ?

[Ken Lee]

I have made some grey-scale step wedges, and notice that as I move the mouse over the different steps, I see different brightness values within the same region.

Is this a common occurrence ? I would think that if PS designates a region as, say 50% brightness, or RGB 122, 122, 122, then the entire region would be the same value.

Perhaps someone could try this and tell me if you get the same results.

I am using CS3 on a Mac Powerbook G4.

It's hard to calibrate something against a step wedge, if the step-wedge itself is non-uniform.

[Colin Graham]

I had this problem making a 101 step tablet and ended up doing it in long hand; i.e. making a grid and using paintbucket to fill each one with the correct K% and this gives perfectly smooth steps. I had tried the gradient fill method with adjust>posterize or the pixelate>mosaic filter and got results like yours. If you'd like a copy of the high res version don't hesitate to ask.

[Ken Lee]

Colin -

On Windows + Photoshop CS, my JPG step wedge appears entirely uniform within each step. One or two of the values are off by 1 point, but I get the same value wherever I move the mouse within a step.

On the same platform, looking at your step wedge, I also get perfectly uniform values along the top row, but the values themselves are surprising: Starting at the left, they are as follows: 0, 1, 2, 4, 6, 8, 9, 11, 12, 14. If I view the same image on my Mac with CS3, the values are not uniform: they vary widely inside each step.

Attached is a simple tablet with Brightness values of 0, 25, 50, 75, and 100. On saving to JPG, the 0 changed to 1. I made it on Windows, with Photoshop CS. If I view that same file on the Mac with CS3, the values in each step appear to vary.

I hope that this isn't a Mac problem, but rather a PS problem. Maybe we have found a long-standing bug. Either the values are variable, and we have just learned about it, or the values are uniform, and either CS3 or Mac can detect the difference.

I would love to know which which, is which.


Here

[Marko]

Ken,

I think your problem is in the output format - jpg - rather than Windows.

Jpeg is a lossy, compressed format suitable for images with lots of shades and continual transitions such as gradients.

Gif, on the other hand, is a losless format suitable for large, uniform areas of solid color.

As a rule of thumb, it is best to use jpeg for photos and gifs for illustrations and diagrams.

I hope this helps.

[Brian Ellis]

If you don't get a satisfactory answer here you might post your question to the Yahoo group "digitalblackandwhite:theprint." There are questions and anwers there all the time about using step wedges for linearizing and other purposes.

[Marko]

One more thing - The best way to save it as a gif would be to use Save for Web and then pick gif - 256 colors - no dither. This should give you perfectly uniform values for up to 256 shades of gray.

[Ken Lee]

"I think your problem is in the output format - jpg - rather than Windows"

Thanks Marko, I thought about that too, but the non-uniform values are apparent before the images are saved.

Just start with a greyscale image, select a region, and "pour" some value into it with the paint bucket tool. Move the mouse around, and you will see (on Mac + CS3) that the values fluctuate considerably. Try the same procedure on Windows + CS, and the region appears uniform. Don't save anything.

[Ken Lee]

"you might post your question to the Yahoo group "digitalblackandwhite:theprint."

Thanks Brian.

After a quick glance, it seems that many people on that site, use the step wedges which ship with the Quadtone Rip. Those are in PSD format.

I had a look at them on my Mac + CS3 setup. The values fluctuate, just as they do when I look at my own step wedges. On my Windows setup, they are perfectly uniform.

So again, I wonder: is it something inherent on the Mac, or in CS3 ?

[Marko]

Ken,

I just tried several methods on CS3/Mac and was not able to reproduce your problem. I tried the following:

1. Black to White gradient edge to edge followed by Posterize/21 steps
2. Rectangular selection then paint bucket
3. Rectangular selection then Option/Delete

All three methods resulted in perfectly even patches.

May I suggest you try again and make detailed notes so we can try to replicate the problem? If it is a bug, it would be useful to find out and report it, and if the problem is entirely yours, then you will end up solving it... :)

[Ken Lee]

Wonderful -

I will do that. I have to run out now for a few hours, but will do so as soon as I can. Note that Colin says he had the same problem. Perhaps he can explain his method too.

Many many many thanks !

[Colin Graham]

In trying to build some curves for alt printing I've been making some step tablets, and at first I tried using the fill gradient black to white and then posterized 21 steps or used the pixelate-mosaic filter with around a 78 cell size for a 5" long strip. The result were steps that varied within each step, almost like mini-gradients within each step, that look a lot like the example you posted Ken. I'm running an ancient version of ps v6 on windows XP, so that might be something there.

Using the fill tool for each square individually seemed to cure the variance, at least with my workstation. Possibly the odd sequencing you're getting from my example is a result of the web-optimized file? I didn't want to try and post the full 3mb jpeg. But it is odd that you're only getting the in-the-step variance problem on a mac.

Anyway, my digital negative excursion had been stalled on getting a decent printable step tablet so I can plot some curves for my process; I thought you may have been in similar straits. Sorry if I misunderstood your original question. Cheers, Colin

[Marko]

Colin,

You got me interested enough to run a few experiments and I found that the key to even values in the patch is to run the gradient with no dither. It's been a while since I used PS 6, but in CS3, there is a context-sensitive toolbar at the top that changes depending on the tool you have selected.

When you select gradient, you will see three checkboxes among all the other options in the toolbar: Reverse, Dither and Transparency. Just make sure that you have all three deselected.

Next, make sure that you have pure black and white selected as your foreground and background color - just hit "D" on your keyboard (case doesn't matter) and you'll have it set.

Also, make sure you have foreground-to-background gradient selected and not foreground-to-transparent.

Then hold down Shift and drag your cursor from the left edge to the right. (Holding Shift makes it perfectly straight).

Finally, run the Pixelate/Mosaic. Now, here's the tricky part: in CS3 the dialog lets you choose the size of the box in pixels but then produces only as many evenly-sized blocks as your document width allows.

So, you have to plan in advance and set your image width as the N x W px, where N = number of steps (in this case 21) and W = cell width in pixels. W is the value you need to enter in the Mosaic dialog.

Say you want to create a 21-step wedge, with each step 20px wide. You would create 420px wide document (pick your height as you please) and then enter 20 into the Mosaic dialog.

Another aspect to pay attention to is that in doing this, you will need to plan your pixel dimensions according to the resolution you will be printing it at. The reason being that your step wedge is a raster and not vector, so it does not lend itself well to resizing.

This is a quck wedge I created using this method and quoted values:

4915

Saved with Save for Web - 32-color gif (what matters is to have more colors than wedges) - no dither (again) - no transparency.

[Ken Lee]

Marko - On my machine, things still behave wrongly when I view that image.

I ran over to the local Apple Store, and tried out a new Intel-based laptop with CS3 - and everything works fine. So either there is something wrong with my setup, or there is a flaw inherent in my hardware configuration.

I will uninstall and reinstall CS3. Perhaps that will correct the problem.

[Marko]

Ken, hold on with reinstalling...

You are right, my image displayed the same thing too when viewed over the Web (saved through the browser).

So I ran a few more tests...

[Marko]

Ok, these are my notes:

1. Created two 21-step greywedge files using the above described method:

a) 8-bit RGB
b) 8-bit Grayscale

Pixel-peeping at max. enlargement revealed no pixelation, only perfectly solid blocks.

2. Saved both files using Save As... Compuserve GIF with the following params:

a) RGB

Palette: Exact
Colors: 21
Forced: None
Transparency: None
Row Order: Normal

b) Grayscale

Row Order: Normal

3. Saved both files using Save for Web (and Devices) with the following params:

128-color GIF
Dithering: No
Transparency: No

4. Pixel-Peeping at max. enlargement - and here is where things get interesting - revelead the following:

a) Indexed file (file opened as saved, no reverse conversion): No pixelation
b) Converted all four files to 8-bit RGB: No pixelation
c) Converted all four files to 8-bit Grayscale: Pixelation!

So, it apears that everything is going fine until converting the indexed file to Grayscale.

In other words, neither RGB nor Grayscale file showed any pixelation in native (PSD) form nor in Indexed (GIF) form, so the anomaly appears to be connected to conversion from Indexed to Grayscale rendering.

P.S. After writing this last paragraph, I did a quick Save As TIFF with no compression on both files and there was no pixelation either.

[Colin Graham]

In retracing my steps I found something interesting. Remaking a 21-step wedge, in both posterize and pixelate, the eyedropper claims the steps to be consistant but as far as WYSIWYG and actual printer output go, they aren't at all. After scanning the printed step tablet back into ps the values are all over the place. This is perhaps a profiling error on my end... Anyway, uncle! This is odd though because my monitor and printer output match up very very well and the 101 step version I posted scans back in just like the original file.

Marko- just read your suggestions and will try them out directly. Thanks very much.

Just made this- looks much much better on my end. No pixels at 1600% and very smooth. Thanks for the advice. I also kept forgetting to manually enter the black and white points, mine always seem be at 98% and 2% by default.

[Ken Lee]

"...when viewed over the Web (saved through the browser)"

Marko - You are very kind to go to such lengths of investigation.

I'm suspect that my problem has nothing to do with saving the file in one format or another. All of the files I have mentioned, look fine when viewed either on my XP machine running CS, or on an Intel Mac at the store running CS3.

Logic suggests that I have changed some kind of setting, or there is something inherent in my PowerPC G4, which makes these same files appear to be non-uniform.

The area of embedded profiles and soft-proofing introduces another layer of potential problems. After re-installing, I will see what the "defaults" are. PS doesn't really let you set the whole app back to default settings.

[Marko]

Marko - You are very kind to go to such lengths of investigation.

Ken, not at all, this is my hobby, I do it because I like it. I find experimentation and solving problems much more interesting than watching TV or something equally dull. :)



I'm suspect that my problem has nothing to do with saving the file in one format or another. All of the files I have mentioned, look fine when viewed either on my XP machine running CS, or on an Intel Mac at the store running CS3.

If I understand this correctly, you are saying that the very same files that look fine on other computers look pixelated when viewed on your Mac. Is that correct? Are we talking "same" as in physically same PSD files or just "same" as in the same kind of files?



Logic suggests that I have changed some kind of setting, or there is something inherent in my PowerPC G4, which makes these same files appear to be non-uniform.

Somehow, it all comes back to dithering somewhere in the chain. I'd start with your graphics card bit-depth (number of colors) and work my way up from there before I go radical. But that's just me. Please let us know what do you find out? :)

[Ken Lee]

I do it because I like it. I find experimentation and solving problems much more interesting than watching TV or something equally dull. :)
My good fortune.

If I understand this correctly, you are saying that the very same files that look fine on other computers look pixelated when viewed on your Mac. Is that correct?
I'm not talking about anything that is visible. I mean that the INFO tool, when configured to give HSB, RGB, etc. reports different numbers, as the mouse is moved around within a given region.

Are we talking "same" as in physically same PSD files or just "same" as in the same kind of files?
The same physical files. I create a new file in PS, draw the gradient, then choose the posterize option. I don't save any file. I just mouse over the regions, and notice that within a given region, the values differ in the INFO tool. It all LOOKS fine. If I save the file and reopen it, I see the same thing - unless I open the same file on a different machine. In that case, everything looks perfect. All the INFO numbers are uniform.

Now I just went and uninstalled CS3, and reinstalled my old CS. It also suffers from this problem. So it appears that the problem is due to something on my PowerBook G4. Perhaps a different implementation of some low-level service has some rounding errors that crop up. Ah... computers !

[Ken Lee]

Marko - What kind of machine are you using for OS X ?

Laptop or Desktop ?

Is it an Intel or Power PC machine ?

Just trying to pin this down.

[Marko]

Marko - What kind of machine are you using for OS X ?

Laptop or Desktop ?

Is it an Intel or Power PC machine ?

Just trying to pin this down.

Mostly iMac and PowerBook Pro, both Intel.

But again, I don't think it would be the CPU issue, it sounds more like graphics. I had the G5 tower and the G4 QuickSilver before and never noticed something like this. Unfortunately, I don't have either of those any more, but what I can do is run a quick test on a high-end late model G5 tower next time I go see one of my friends who still have them. I'll let you know what I find.