Luminous Landscape Forum

Raw & Post Processing, Printing => Digital Image Processing => Topic started by: 32BT on February 07, 2013, 05:25:53 am

Title: From the Book of Rules on Image Processing
Post by: 32BT on February 07, 2013, 05:25:53 am

Rule #1:
Thou shalt optimise global contrast first and foremost, before anything else, especially and most importantly before even considering local contrast.

Failure to comply with rule #1, even under the guise of "artistic intent" will make eyes bleed, either your own or those of others.
Title: Re: From the Book of Rules on Image Processing
Post by: Bart_van_der_Wolf on February 07, 2013, 06:12:56 am
Rule #1:
Thou shalt optimise global contrast first and foremost, before anything else, especially and most importantly before even considering local contrast.

Failure to comply with rule #1, even under the guise of "artistic intent" will make eyes bleed, either your own or those of others.

Hi Oscar,

That of course only applies when the image is not converted to a floating point version first, rule #0.

Using a parametric approach which wraps as many adjustments as possible into a single one, also helps, but is not as good as floating point calculations from the start to end.

Cheers,
Bart
Title: Re: From the Book of Rules on Image Processing
Post by: 32BT on February 07, 2013, 06:36:53 am
That of course only applies when the image is not converted to a floating point version first, rule #0.

Nope, it's valid in each and every case. By Global Contrast I mean that first and foremost one should distribute the tones over the available range for output. Only if you have exhausted the possibilities and range of that distribution does one resort to Local Contrast adjustments.

The problem with Local Contrast adjustments is that they are size dependent and therefore introduce all kinds of funky looking transitions if viewed in a different size than originally intended or used. What may look good on a 30" wide gamut display with one's nose 2 inches away from the screen, generally falls apart when published on the internuts.

Perhaps that would be the rule of thumb for local contrast adjustment: the transition should be invisible (or realistic) regardless of imagesize.
Title: Re: From the Book of Rules on Image Processing
Post by: Tim Lookingbill on February 07, 2013, 12:42:29 pm
Nope, it's valid in each and every case. By Global Contrast I mean that first and foremost one should distribute the tones over the available range for output. Only if you have exhausted the possibilities and range of that distribution does one resort to Local Contrast adjustments.

The problem with Local Contrast adjustments is that they are size dependent and therefore introduce all kinds of funky looking transitions if viewed in a different size than originally intended or used. What may look good on a 30" wide gamut display with one's nose 2 inches away from the screen, generally falls apart when published on the internuts.

Perhaps that would be the rule of thumb for local contrast adjustment: the transition should be invisible (or realistic) regardless of imagesize.

Can you show us an example of what you're seeing?

I experience something similar but only editing in Photoshop. Adjusting contrast on a Raw image (and sometimes on gamma encoded jpegs/tiffs) in ACR set to 16 bit/ProPhotoRGB output space is a much more smoother and refined experience.

I also find adjusting contrast in ACR works in concert with Clarity tweaking little by little back and forth with regard to controlling the appearance of tonal edge roll off without noticeable halos. If things get too saturated usually a tweak to the WB and sometimes saturation slider fixes it but all this only applies to shots taken under mixed lighting and varied contrasted scenes.

Controlled studio lighting shooting environments act more like working in a copy stand setup where edit one image once, copy and apply to the rest and you're pretty much done.
Title: Re: From the Book of Rules on Image Processing
Post by: Fine_Art on February 07, 2013, 01:21:33 pm
Nope, it's valid in each and every case. By Global Contrast I mean that first and foremost one should distribute the tones over the available range for output. Only if you have exhausted the possibilities and range of that distribution does one resort to Local Contrast adjustments.

The problem with Local Contrast adjustments is that they are size dependent and therefore introduce all kinds of funky looking transitions if viewed in a different size than originally intended or used. What may look good on a 30" wide gamut display with one's nose 2 inches away from the screen, generally falls apart when published on the internuts.

Perhaps that would be the rule of thumb for local contrast adjustment: the transition should be invisible (or realistic) regardless of imagesize.

I think Bart is right.

The very first thing you should do if your software can handle it, is convert the camera's raw into 32 bit FP. Any data manipulations will be performed with greater precision. I have tried very strong curves, saturation changes, color blend shifts, then reversed them in the same order they were initially done. I can see no degradation in the histogram. The image also looks good. I save in 32 bit fit. Saving to 16 bit tif from my master also leaves me with huge flexibility.
Title: Re: From the Book of Rules on Image Processing
Post by: 32BT on February 07, 2013, 01:24:01 pm
Can you show us an example of what you're seeing?
...
Controlled studio lighting shooting environments act more like working in a copy stand setup where edit one image once, copy and apply to the rest and you're pretty much done.

Well, an excellent example (but just one of many posted here recently) would be this:

http://www.luminous-landscape.com/forum/index.php?topic=74837.0 (http://www.luminous-landscape.com/forum/index.php?topic=74837.0)

If you crop the frame off and then look at the histogram it will be immediately obvious that the maximum range isn't even close to being fully used. But it is also immediately clear that halo's are present. So, the image should first be rendered for global contrast to fully utilise the range, then one can opt to do additional local contrast enhancement to emphasise the sky and whatever else.

And in that second step one should watch for the creation of visible halo's, just like you mentioned, and that may involve a back-and-forth process like you mention also, but it is probably most easily seen by looking at a smaller sized version of the file under edit. If the halo's show in the smaller size, then it is most likely unrealistic and/or incorrect and generally unpleasant to look at.

And, yes, images from controlled studio lighting will likely never exhibit this type of problem. Or perhaps under some special circumstances the halo's will be introduced by intent to emphasise depth or achieve a certain look. Then of course, the halo size will be specifically adapted to the subject. (The well known Balkan style HDR portraits come to mind).


Title: Re: From the Book of Rules on Image Processing
Post by: 32BT on February 07, 2013, 01:31:10 pm
I think Bart is right.

The very first thing you should do if your software can handle it, is convert the camera's raw into 32 bit FP. Any data manipulations will be performed with greater precision. I have tried very strong curves, saturation changes, color blend shifts, then reversed them in the same order they were initially done. I can see no degradation in the histogram. The image also looks good. I save in 32 bit fit. Saving to 16 bit tif from my master also leaves me with huge flexibility.

Yes, greater bit depth is useful to have sometimes, but the greater bit depth doesn't stop people from underutilizing the dynamic range and introducing halo's by applying unrealistic local contrast adjustments.

Title: Re: From the Book of Rules on Image Processing
Post by: Tim Lookingbill on February 07, 2013, 02:00:52 pm
Well, an excellent example (but just one of many posted here recently) would be this:

http://www.luminous-landscape.com/forum/index.php?topic=74837.0 (http://www.luminous-landscape.com/forum/index.php?topic=74837.0)

If you crop the frame off and then look at the histogram it will be immediately obvious that the maximum range isn't even close to being fully used. But it is also immediately clear that halo's are present. So, the image should first be rendered for global contrast to fully utilise the range, then one can opt to do additional local contrast enhancement to emphasise the sky and whatever else.

And in that second step one should watch for the creation of visible halo's, just like you mentioned, and that may involve a back-and-forth process like you mention also, but it is probably most easily seen by looking at a smaller sized version of the file under edit. If the halo's show in the smaller size, then it is most likely unrealistic and/or incorrect and generally unpleasant to look at.

And, yes, images from controlled studio lighting will likely never exhibit this type of problem. Or perhaps under some special circumstances the halo's will be introduced by intent to emphasise depth or achieve a certain look. Then of course, the halo size will be specifically adapted to the subject. (The well known Balkan style HDR portraits come to mind).




All I see in your linked thread is a black and white image of Yosemite Valley.

Not exactly making my eyes bleed as you pointed out. There's no before and after either.

I have quite a few Raws I've shot and "restored" and with some applied "color grading" effects where contrast distortions abound but with nothing I would deem as undesirable results. No posterizing. No unwanted artifacts. No halos (noticeable at least) and my eyes weren't bleeding. I got exactly what I wanted.

Have another example? Color?
Title: Re: From the Book of Rules on Image Processing
Post by: 32BT on February 07, 2013, 02:20:11 pm
Have another example? Color?


http://www.luminous-landscape.com/forum/index.php?topic=69836.0 (http://www.luminous-landscape.com/forum/index.php?topic=69836.0)


http://www.luminous-landscape.com/forum/index.php?topic=74794.0 (http://www.luminous-landscape.com/forum/index.php?topic=74794.0)
Title: Re: From the Book of Rules on Image Processing
Post by: Fine_Art on February 07, 2013, 03:42:22 pm

http://www.luminous-landscape.com/forum/index.php?topic=69836.0 (http://www.luminous-landscape.com/forum/index.php?topic=69836.0)


http://www.luminous-landscape.com/forum/index.php?topic=74794.0 (http://www.luminous-landscape.com/forum/index.php?topic=74794.0)

You have lost your mind if you say Matthiew's ebook is an example of badly done contrast. He uses a full range of black to white where appropriate. Some images simply don't have blacks. Those images are well done.
Title: Re: From the Book of Rules on Image Processing
Post by: bill t. on February 07, 2013, 06:07:54 pm
That of course only applies when the image is not converted to a floating point version first, rule #0.

Humh.  So do I not get the advantages of casting integer raw files to floating point if I do not first convert them to say dng files, but rather keep all my camera files only as raws (using LR)?  My panos sometimes generate 100+ camera files per image, and I have avoided dng's and such simply to keep down the file glut.  Must confess to some ignorance in this area.
Title: Re: From the Book of Rules on Image Processing
Post by: Tim Lookingbill on February 07, 2013, 06:21:33 pm

http://www.luminous-landscape.com/forum/index.php?topic=69836.0 (http://www.luminous-landscape.com/forum/index.php?topic=69836.0)


http://www.luminous-landscape.com/forum/index.php?topic=74794.0 (http://www.luminous-landscape.com/forum/index.php?topic=74794.0)

The first one was enough to get your point across. The second link downloading the ebook pdf took way too long to download with no progression indicator on how long it would take. I gave up.

Seirraman's Bryce mesa shot suffers somewhat from a lack of contrast in relation to appearance of saturation considering the blue sky canopy de-saturation effect in a sunset scene. There's also far less definition in the foreground green shrubbery shadow area in relation to too much definition and detail in the distant hills and mesas which fights with the optics we humans tend to associate with near and far objects. Learned that in landscape painting 101 in high school.

Over cranking Fill slider can give the over saturation/low contrast effect. Editing in Lab tends to make images look like that as well for those that love adding saturation without changes to luminance which isn't how the sun works reflecting on real objects especially those viewed in overcast lighting.

It's still an overall pleasant looking image. Just need to dial back the sharpness so the level of detail recedes the farther into the distance. Some like to apply sharpness masking and painting it back in to get this effect.

But we don't know if the photographer adjusted contrast in the order indicated in your rules you stated. Looks like he got overzealous in his sharpening edits finding that a digital camera can capture far more than our eyes can see when viewing that much detail in a high rez image at 100%. I'm guilty as charged of it as well.