8 bit VS 16 bit in editing

[LittleLew]

The following figure might give you some hint of why I like 16 editing.
All these are clips of the levels histogram.
The top row is a 16 bit image before and after levelling adjustment. Note the smooth curve. I interpret this as a smooth transition of tones from one place on the intensity scale to another.

The bottom role is the same image - after conversion to 8 bit.
Note that the post-level image is rather irregular with non-intuitive (to me at least) peaks and valleys. I interpret that to mean that since I have discarded some data, the tone conversion is not as neat and smooth as it should be.

(I welcome any better and more knowledgeable interpretation.)

[arodney]

Originally Posted by LittleLew

I interpret that to mean that since I have discarded some data, the tone conversion is not as neat and smooth as it should be.
(I welcome any better and more knowledgeable interpretation.)

The important indicators here are the white "lines" within the Histogram which represent data loss. The height of a particular black line only indicates the number of pixels at that level within the image. Not really important.

Note that there are also cases that are equally damaging that do not show up in a Histogram. You're seeing the effect of stretching the tone curve, like a rubber band. Values are being lost as indicated by the white lines. The opposite effect can happen when you take two different but close values and, with an edit, combine them into one (this wouldn't show up in the Histogram). For example:

http://staging.digitalphotopro.com/t...decision.html:

Suppose you build a smooth linear gradient in Photoshop from black to white in 8-bit and use one pixel for each value. You’d have 256 tones. Level 0 is used to represent pure black, Level 1 represents the value nearest to that black but just a bit lighter, and so on, going all the way to a pure white at Level 255. The Photoshop Histogram plots these values using a single black line to represent each of the 256 values running left (Level 0) to right (Level 255).

Now you apply a simple edit in Photoshop, say you pull a curve to lighten the image. You change many of the numeric values and their relationship to each other. Multiple and differing levels are now represented by a single value while individual values may be discarded. For example, Levels 1, 2 and 3 are all now defined as Level 1. Levels 128, 129, 130 and 131 may now be represented by a single value, perhaps Level 131.

[LittleLew]

Suppose you build a smooth linear gradient in Photoshop from black to white in 8-bit and use one pixel for each value. You’d have 256 tones. Level 0 is used to represent pure black, Level 1 represents the value nearest to that black but just a bit lighter, and so on, going all the way to a pure white at Level 255. The Photoshop Histogram plots these values using a single black line to represent each of the 256 values running left (Level 0) to right (Level 255).

Now you apply a simple edit in Photoshop, say you pull a curve to lighten the image. You change many of the numeric values and their relationship to each other. Multiple and differing levels are now represented by a single value while individual values may be discarded. For example, Levels 1, 2 and 3 are all now defined as Level 1. Levels 128, 129, 130 and 131 may now be represented by a single value, perhaps Level 131.

This seems to be a good arguement for shooting in RAW rather than in JPG , an inherently lossy format.

[arodney]

Originally Posted by LittleLew

This seems to be a good arguement for shooting in RAW rather than in JPG , an inherently lossy format.

That's one but there are way more reasons to be shooting Raw:
http://tinyurl.com/33msxz

In another thread here, there's talk of all this LAB color correction work and I have to keep asking myself, why are all these awful images ending up in Photoshop for this work in the first place? Why are people (at least those making up these techniques) not getting the best possible color rendering from the get go from Raw? Its not easy to get the toothpaste back into the tube when you start with bad images. With proper Raw rendering, the need to "fix" any kind of global color and tone work in Photoshop should be about nill. But that's for another post.....

[Dogdots]

Originally Posted by LittleLew

The following figure might give you some hint of why I like 16 editing.
All these are clips of the levels histogram.
The top row is a 16 bit image before and after levelling adjustment. Note the smooth curve. I interpret this as a smooth transition of tones from one place on the intensity scale to another.

The bottom role is the same image - after conversion to 8 bit.
Note that the post-level image is rather irregular with non-intuitive (to me at least) peaks and valleys. I interpret that to mean that since I have discarded some data, the tone conversion is not as neat and smooth as it should be.

(I welcome any better and more knowledgeable interpretation.)

I found this very interesting. It's good to see the reason for things visually

[Duffy Pratt]

Andrew:

ACRs new capabilities are great, and I'm now using it for most of my RAW work. There are still some things that I find much easier to do with Photoshop, in RGB, LAB, or even sometimes in CMYK.

My wife's P&S is .jpg only. For shots she has taken, I'm doing much more work in Photoshop, and resort to LAB with more frequency.

So for me the answer to your question is because, for some tasks, LAB is easier and quicker. And for some shots, ACR just isn't a real option (I don't think its very useful for starting on JPGs).

Duffy

[arodney]

Originally Posted by Duffy Pratt

There are still some things that I find much easier to do with Photoshop, in RGB, LAB, or even sometimes in CMYK.

So for me the answer to your question is because, for some tasks, LAB is easier and quicker. And for some shots, ACR just isn't a real option (I don't think its very useful for starting on JPGs).

Can you provide some examples?

[Duffy Pratt]

I don't have any means for uploading RAW files now. So I won't be able to show an example that also gives a way to prove that what I've done can be done easier or better with Camera Raw. If you really want, I will try to come up with some images where I still resort to LAB.

I can offer some general discriptions:

I still sometimes prefer writing A and B curves to boost color contrast, especially in landscapes. This becomes even more attractive when I quickly want to boost some colors, while locking down others. For example, if I need to lock down some bright yellow flowers, but want to perhaps cut some yellow in skin tones, I can sometimes do this quickly in LAB and get a result that is acceptable. Frankly, this issue doesn't happen all that often anymore with ACR.

On portraits, I will sometimes use LAB to apply the A and B channels to themselves in overlay mode. I haven't found a faster way to get this effect for making faces more lively.

Here's another issue I can solve readily with LAB. I have white dogs. At sunset, they often appear blue in shadows and warm where they are directly lit. White balance in ACR sometimes cures this problem, but not always. With LAB, I can sometimes leave some Blue in the coat, and then eliminate it with a curve and a "blend if" filter. I've tried doing the same thing in ACR with the HSL sliders, and I can't get it to work as well for me.

Edit addition: another LAB trick I've used recently is doing a high resolution, low amount sharpen on the A and B channels, through a luminosity mask.

I know some images would help alot, and I'll try to look and find the best ones. Also, please note that I haven't claimed that LAB is definitively better for all these examples -- just that I've found it easier, and sometimes better, given the current state of my knowledge of these tools. I'm always happy to learn better or easier ways to get things done. (That said, I have a mathematics background, and generally tend to reduce everything to a problem that I've already solved. Thus, I have a natural tendency to stick with tools I've already learned.)

Duffy

[arodney]

Originally Posted by Duffy Pratt

I don't have any means for uploading RAW files now. So I won't be able to show an example that also gives a way to prove that what I've done can be done easier or better with Camera Raw.

At some point, that would be a fun and useful exercise for some of us to try.

I still sometimes prefer writing A and B curves to boost color contrast, especially in landscapes. This becomes even more attractive when I quickly want to boost some colors, while locking down others.

Play with the HSL controls. Very powerful. Also, since you're never burning in pixels, you can build as many variations (Virtual Copies) or, Snapshots as you wish. This is however global in nature just like everything currently in LR or ACR.

Here's another issue I can solve readily with LAB. I have white dogs. At sunset, they often appear blue in shadows and warm where they are directly lit. White balance in ACR sometimes cures this problem, but not always.

Agreed but WB here is a pretty rough edit. Again, HSL might do the trick, at least getting you 90% or more there.

Edit addition: another LAB trick I've used recently is doing a high resolution, low amount sharpen on the A and B channels, through a luminosity mask.

Did you try playing with the Alt/Option key on the capture sharpening sliders in ACR/LR to "build a mask"? I don't do much other sharpening here until output sharpening for print is a possibility. All output sharpening I do is done in the original color mode (using PhotoKit Sharpener).

Also, please note that I haven't claimed that LAB is definitively better for all these examples -- just that I've found it easier, and sometimes better, given the current state of my knowledge of these tools.

Understood. My point, directed more at the book we are discussing is the often awful quality of the images that are "fixed" (not tweaked as we are discussing) and how or why one could need to use such heavy handed, slow and image damaging practices instead of at the pixel creation stage, with the Raw converter.

As a few examples in the book in which giving the Raw's, I'm fairly certain these "fixes" could easily be done in the converter:

Fig 3.1- 3.11(Clarity and Vibrance).

Fig 3.12-3.16 (HLS)

Fig 4.1 4.3 (proper WB, that's easy). Then clarity and Vibrance. Here's an example of truly but-ugly originals of which I speak. Where on earth did they come from? Someone messed up the scan or the Raw conversion.

You get the idea.

[Duffy Pratt]

I think most of the "butt ugly" originals, as you put it, come straight out of the camera as a .jpg. That's the sort of thing I find myself correcting fairly frequently from shots with my wife's P&S. (A Canon G9 will solve that problem for me.)

Shoot some high ISO pictures with a Canon body, using auto white balance, and save as an in camera .jpg. If the shot has anything near the full dynamic range, the only quick way I have found to get rid of the orange cast is using some LAB curves.

I'm not fully persuaded that vibrance and clarity will do the same thing as the LAB techniques. Clarity, from what I can see, looks like it is a local contrast adjustment. Vibrance strikes me as being similar to steepening the A and B channels, but of course it doesn't have the ability to put in control points, and it doesn't separate Blue/Yellow from Green/Magenta. That said, I have found that I'm relying on Vibrance for getting effects that I formerly would have gotten with AB steepening, but there are times when I want more control.

Duffy

[arodney]

Originally Posted by Duffy Pratt

I think most of the "butt ugly" originals, as you put it, come straight out of the camera as a .jpg. That's the sort of thing I find myself correcting fairly frequently from shots with my wife's P&S. (A Canon G9 will solve that problem for me.)

Ah so that explains the kind of images and photographers the book is aimed towards.

But Danno says he gets lots of images from Pro photographers (those poor soles). My wife gets vastly superior images out of her PHD camera in JPEG.

But I guess it makes sense to target these kinds of edits using a $700 editor towards point and shot JPEG capture. OK, I'm of course not being serious.

I'm not fully persuaded that vibrance and clarity will do the same thing as the LAB techniques.

I'm not implying it will produce identical results. The question is, does it do a sufficiently good job of rendering the image as opposed to polishing turds?

Clarity, from what I can see, looks like it is a local contrast adjustment.

Pretty much yes.

Vibrance strikes me as being similar to steepening the A and B channels, but of course it doesn't have the ability to put in control points, and it doesn't separate Blue/Yellow from Green/Magenta.

Its a non linear saturation adjustment with skin tone protection.

The question is, do you need points? Can you use the controls to produce the desired color appearance? Is it less damaging? Is it faster? Does it provide more options? In some cases the answers may be yes or not. The main question remains; when is it best to do as much global image rendering (pixel creation) in the Raw processor versus trying to fix rendered pixels in Photoshop? And that's not solely aimed at LAB edits!

That said, I have found that I'm relying on Vibrance for getting effects that I formerly would have gotten with AB steepening, but there are times when I want more control.

OK, that's useful to know and discuss.