PP Upgrade Conundrum

[Stagecoach]

I have been for many years and still am a user of Elements Version 7.

The only specific areas of concern with my PP work are as follow;

a) The occasional banding I get in certain skies that limits how far I can push them (if I want to).
b) The banding I see in a 'Gradient' mask as per this example.



I had assumed as read from certain articles these situations were due to me working in 8 bit rather than 16 bit.

For these reasons a) and b) only I have recently undergone evaluation of Elements 12 (yes I know still 8 bit), LR5 and finally tonight downloaded PSCC trial.

To my surprise even in PSCC and in 16 bit I get exactly the same banding in gradients and the results when pushing the sky I have used as a trial seem to be no better.

Having downloaded CC tonight, almost falling off my chair in excitement as I see the screen is almost identical to Elements 7 (unlike ver 12 that has gigantic child friendly tool icons) could it be my monitor?

So as a first step to this puzzle, does someone with a 'good' monitor see banding in the above gradient?


Grahame

[Black Pearl]

Yes - but you could also question if that because of the compression for the web.

[Stagecoach]

Yes - but you could also question if that because of the compression for the web.

Robin, the banding clearly shows on a full size image 4000px wide (new file white background) with the gradient added whilst open in editor or on a full size image mask and before downsizing for the web.

[pnodrog]

The problem is that banding can be caused for several different reasons and at different points in the conversion to be displayed or printed.

Jpeg quality is often the main culprit.
Display performance or dyes/pigments used in printing another
Colour space in use and the colours used to generate the gradient that displays banding.
The rendering intents such as absolute and relative colorimetric, perceptual, and saturation.
I suspect 8 or 16 bit channels overall probably has the least effect unless you own a 10bit display.

P.S. I have a reasonable monitor and can see faint banding in your sample but they may not be at the same positions you are observing...

The best you can do is use 16bits per channel in the biggest colour space you have available (prophoto RGB?) and cross your fingers and toes regarding the output..

[IzzieK]

Yes, I can see banding in my monitor on this image ... have you thought of changing your work space to photography instead of the default "Essential" and also changing your Proofing option? Sometimes the simplest one is the most forgotten option...just a thought. It might not even make a difference.

[ajohnw]

As they were rather large I have posted some 65% quality jpg's on here recently. Banding ??? As I understand things this is a typical figure for frugal web use to keep the bandwidth down. I usually use 95%. I wouldn't personally try to do further work on a 65% quality jpg. 95% is a different matter within limits.

I would say the problem is more likely down to raw conversion especially if working in 8 bit later on. It important to get all of the information in the raw file into the colour space in a reasonably acceptable fashion and then augment it with further PP rather than doing the conversion willy nilly and trying to sort everything out later on.

Colin expressed this in a different way. "It's important to leave space for further work". What that means effectively is that when needed all of the histogram space shouldn't be used up - space has to be left for further adjustments which means anticipating what the next stages will be.

The other reason for banding is simply pushing things too far no matter what bit depth is being used. This comes back in some ways to exposures and luminance range in the image particularly problems in things like skies as far as banding is concerned. The dark end tends to have other problems if pushed too far - usually colour artefacts.

It's not a problem I have ever had but I am fairly adept on the use of levels / curves / white points / black points early on even if no one likes my results. I seldom need to use exposure compensation. Colin's tip as obvious as it is has helped me a lot as the usual tendency is to use up all of the space initially. I now have to think about what I may need to do next. Not a problem on mid tones but can be at the dark and bright end if space is not left.

The big advantage of working in deeper colour spaces than the monitor can show really concerns information loss. Using the deeper space means that the fractional bits as far as the monitor is concerned are still there so can be manipulated further. If monitor and workspace are the same once adjusted out they have gone. It's as simple as that.

John
-

[ajohnw]

If people haven't calibrated their monitor it might pay to look at this. The grey scale steps are rather small so any none linearity may show up. Not much chance but this area can cause banding if well out.

http://www.drycreekphoto.com/Learn/C...nsitivity.html

I can just about make out the extremes but even with a screen with very wide viewing angles things are marginally clearer if I look square one.

One thing I forgot to mention about editing at the same bit depth as the screen is that repeated editing of the same thing is more likely to cause problems.

John
-

[chauncey]

Grahame, good calibrated monitor using IE at my usual zoom of 150% banding barely visible, however cranking that
zoom up to 300% really exposes the banding.

[Stagecoach]

Paul, Izzie, John and William,

Thanks for the input.

The problem is that banding can be caused for several different reasons and at different points in the conversion to be displayed or printed.

Fully agree and for that reason things need to be narrowed down to specific areas when it's happening.

P.S. I have a reasonable monitor and can see faint banding in your sample but they may not be at the same positions you are observing...

Very significant point.

The best you can do is use 16bits per channel in the biggest colour space you have available (prophoto RGB?) and cross your fingers and toes regarding the output..

I will change my colour space and see the results.

Yes, I can see banding in my monitor on this image ... have you thought of changing your work space to photography instead of the default "Essential" and also changing your Proofing option? Sometimes the simplest one is the most forgotten option...just a thought. It might not even make a difference.

I need to look further into the set-up of PSCC to investigate this one Izzie.

I would say the problem is more likely down to raw conversion especially if working in 8 bit later on. It important to get all of the information in the raw file into the colour space in a reasonably acceptable fashion and then augment it with further PP rather than doing the conversion willy nilly and trying to sort everything out later on.

This may be applicable to to the 'sky' aspect, if that PP route was taken, but I can not see it being applicable to the gradient problem as described and shown.


As expected there are many variables so to narrow things down I will concentrate on the 'Gradient' banding only first to see if there is a way that I can produce an image as in the example above that is seen on my monitor with no banding (or at least less noticeable).

The criteria I will set will be to make a 'New' file 1200px by 800px in PSCC, white background, using the different options available to see if a smoother gradient can be achieved.

I wonder if anyone using PSCC can actually produce a 1200 x 800 image viewed on their screen at full size that shows absolutely no banding when a gradient is run from top to bottom

PS. 95% of the time the consequences of this banding on a gradient are totally insignificant (read un-noticeable) to me on my finished image where gradients have been used. It is the 5% I'm interested in to evaluate if PSCC is going to improve on this aspect from Elements.

Grahame

[ajohnw]

I've used the GIMP in it's 3x8 bit colour channel form for some time. Recent versions have changed that. Banding has never been a problem but the raw conversion software I used with it was more or less entirely curves driven. That caused me to give that area a lot of attention - getting the image as right as it can be in at that stage.

I also use the GIMP for straight layer work - soft light layers etc. I've never seen that cause banding.

The only time I have had banding is working on other peoples jpg's but not my own even straight out of the camera but this has been working with software that retains more than 8bits per colour channel at the workspace level. I can only assume that other peoples jpg's have gone wrong some where in their work flow and that I suspect has to be initial conversion from raw. Might be bad manipulation, might be down to poor exposure and the image being pushed to far. Some PP packages processes also change the histogram in an odd way. Instead of showing a continuous curve they show a series of "blips" with spaces in between. The tops follow the usual histogram form. This seems to be a characteristic of using say levels in an purely 8 bit channel package and is how banding occurs if it goes too far. I've posted and seen other images that people have produced that show this sort of histogram when loaded into the GIMP but no comments on banding.

One area that might cause problems is trying to work on 8 bit adobe rgb images in an 8 bit colour workspace for aRGB jpg's. The gamut is designed to push the span of the colours covered as far as they realistically can be without undue banding problems while using 8 bit colour channels. This really does need the raw file data retaining all of the way through the work flow. There is another thread up currently concerning workspaces and output gamuts and why there is a need for extra depth in the workspace. Various gamuts in real terms are not directly connected with this need. They just need to hold a higher bit precision than the output = more colours after a fashion that can even be in a different gamut.

Some people set the camera to aRGB and work on the jpg's in sRGB. They like the colour effect it tends to give. As the colours are now much closer together banding is far less likely.

I assume your monitor is calibrated - it's important that it can produce and even grey scale. Probably more important than perfect colour reproduction within limits. The only sure way of getting it to do that or knowing how bad it is if the package allows that is calibration. One software package that will work with most colorimeters and show just how good/bad it is is dispcalgui which can be found on the web.

Off to bed - hope my weird typo's are limited.

John
-

[chauncey]

I've been following this thread and a question come to mind...
what are your goals in this ridiculous hobby of photography?

You invest money in camera/computer/color calibrated monitor gear, learn how to use that gear
then you shoot the baked in, manufacture's jpeg images and then, when you aren't satisfied with
them, it's off to PP to "fix" that unsatisfactory image, often with inadequate results. Why?

Instead, why not take the time to do it right from the outset...shoot in RAW>spend $10/mo to get
PS CC and invest the time for the learning curve.

When you consider the time involved in fixing your screwed up images, $10/mo is chicken feed.

[Stagecoach]

I've used the GIMP in it's 3x8 bit colour channel form for some time. Recent versions have changed that. Banding has never been a problem but the raw conversion software I used with it was more or less entirely curves driven. That caused me to give that area a lot of attention - getting the image as right as it can be in at that stage.

I also use the GIMP for straight layer work - soft light layers etc. I've never seen that cause banding.

The only time I have had banding is working on other peoples jpg's but not my own even straight out of the camera but this has been working with software that retains more than 8bits per colour channel at the workspace level. I can only assume that other peoples jpg's have gone wrong some where in their work flow and that I suspect has to be initial conversion from raw. Might be bad manipulation, might be down to poor exposure and the image being pushed to far. Some PP packages processes also change the histogram in an odd way. Instead of showing a continuous curve they show a series of "blips" with spaces in between. The tops follow the usual histogram form. This seems to be a characteristic of using say levels in an purely 8 bit channel package and is how banding occurs if it goes too far. I've posted and seen other images that people have produced that show this sort of histogram when loaded into the GIMP but no comments on banding.

One area that might cause problems is trying to work on 8 bit adobe rgb images in an 8 bit colour workspace for aRGB jpg's. The gamut is designed to push the span of the colours covered as far as they realistically can be without undue banding problems while using 8 bit colour channels. This really does need the raw file data retaining all of the way through the work flow. There is another thread up currently concerning workspaces and output gamuts and why there is a need for extra depth in the workspace. Various gamuts in real terms are not directly connected with this need. They just need to hold a higher bit precision than the output = more colours after a fashion that can even be in a different gamut.

Some people set the camera to aRGB and work on the jpg's in sRGB. They like the colour effect it tends to give. As the colours are now much closer together banding is far less likely.

I assume your monitor is calibrated - it's important that it can produce and even grey scale. Probably more important than perfect colour reproduction within limits. The only sure way of getting it to do that or knowing how bad it is if the package allows that is calibration. One software package that will work with most colorimeters and show just how good/bad it is is dispcalgui which can be found on the web.

Off to bed - hope my weird typo's are limited.

John
-

John,

Too many ifs, buts, possibilities, assumptions and variables to address my specific problem. Whilst I do not disagree with anything you have said the fact that you may not have come across a problem similar to that I am investigating could very simply be because you are not working on the specific image content I am.

With regard to the specific problem I have encountered when pushing some skies, of which are always the sunset/sunrise muted tone ones I have used one specific image (as a representative sample where this occurs) pushed it in 8bit and then 16bit and am at the present conclusion that 16bit has not allowed me to push it any further.

I still have further testing to do with this specific image using different colour spaces and may find that there is a mode that gives me more scope but it appears at present that for the specific comparison I'm evaluating the the 8/16 bit alternatives are not the cause.

It may very well be that the 'limitation' encountered has nothing to do with the previous restriction I had of only being able to work in 8 bit but simply due to an image not being able to be pushed any farther due to it's content.

Grahame

[Stagecoach]

I've been following this thread and a question come to mind...
what are your goals in this ridiculous hobby of photography?

You invest money in camera/computer/color calibrated monitor gear, learn how to use that gear
then you shoot the baked in, manufacture's jpeg images and then, when you aren't satisfied with
them, it's off to PP to "fix" that unsatisfactory image, often with inadequate results. Why?

Instead, why not take the time to do it right from the outset...shoot in RAW>spend $10/mo to get
PS CC and invest the time for the learning curve.

When you consider the time involved in fixing your screwed up images, $10/mo is chicken feed.

I trust this one was not aimed at me William

I never use a JPEG image for PPing unless on the extremely very rare occasion that I have had to shoot in JPEG due to burst shooting buffer restriction or getting low on card space.

The banding concern with certain sky types is on files that have been shot in RAW taking the absolute utmost care to expose correctly at the highest quality available on the camera. Are there people out there that do not do this

The concerns are not of a criteria that could be termed attempting a 'fix' in PP

As for learning curves of PSCC I amazed myself last night in that after 5 minutes I could do exactly what I would generally do in my very old and capable Elements 7 The most difficult thing I found was that instead of being able to use those two great little buttons with radial CW & ACW arrows as in Elements 7 I have had to learn to press the Ctrl and Z keys at the same time

Grahame

[TonyW]

I just constructed a file with PSCC which is a single colour with a gradient applied. When I saved it as a jpg with maximum quality there was no banding but with minimum quality there was distinct banding. Could this be your problem?

1.Max quality



2. Min quality

[Stagecoach]

I just constructed a file with PSCC which is a single colour with a gradient applied. When I saved it as a jpg with maximum quality there was no banding but with minimum quality there was distinct banding. Could this be your problem?

Afraid not Tony as my concerns/investigations are only with respect to what I see on the screen prior to any re-sizing, compression or saving.

I'm unable to produce an image such as yours at 1200px wide that can be viewed on screen in 8 or 16 bit that does not show banding, or whatever the term for horizontal lines of random width variations in tone is.

[pnodrog]

I just constructed a file with PSCC which is a single colour with a gradient applied. When I saved it as a jpg with maximum quality there was no banding but with minimum quality there was distinct banding. Could this be your problem?

1.Max quality



2. Min quality

Nice colours.

We could start a competition - who can make a gradient of two colours that exhibits the most banding when saved as minimum quality jpeg....

[TonyW]

In that case, I would be interested to know whether you get the same as me if you construct a file with a single colour and then apply a gradient. Do you then see bands before you save it?

I believe that my monitor is a good one (except for a couple of dead pixels) and I can see faint banding in your image. I can't see it at all in my first image. It would be interesting if you could do exactly as I have done to find out if we see the same thing. Do you see my first image without banding?

[pnodrog]

In that case, I would be interested to know whether you get the same as me if you construct a file with a single colour and then apply a gradient. Do you then see bands before you save it?

I believe that my monitor is a good one (except for a couple of dead pixels) and I can see faint banding in your image. I can't see it at all in my first image. It would be interesting if you could do exactly as I have done to find out if we see the same thing. Do you see my first image without banding?

No I your first image seems reasonable free of banding. The images in my post are just part of the quote from your post. Don't try and blame me for them - of course I'll take any credit going....

[xpatUSA]

In the posted image there is most certainly banding but not terribly obtrusive on my cheap uncalibrated monitor. However, ImageJ shows it only too well:



The interesting bit is that the 'steps' are of uneven heights. I'm sure there must be an explanation . . .

[Stagecoach]

In the posted image there is most certainly banding but not terribly obtrusive on my cheap uncalibrated monitor. However, ImageJ shows it only too well:

The interesting bit is that the 'steps' are of uneven heights. I'm sure there must be an explanation . . .

Thanks for that Ted,

I have spent the last hour trying every possible combination of colour-space and settings in PSCC and every one shows banding on the image on-screen to an equal or greater degree than indicated in the image posted. What is common is that the steps as you called them are allways uneven.

I do have limited knowledge of colour space and settings but it 'appears' at present that changing things is not providing a solution.