Dynamic Range: Seeing the Unseen.

[cmr164]

Ok it is my turn to do a mini tutorial. Lots of pictures have 2 distinct ranges of intensities and some (lit face, darkened room, sunlit scene out window) can even have more. John's (RUTT) tutorials on dynamic range and curves are indispensable in using and compensating photos of this nature and this tutorial will not attempt to restate those. Instead I will address the issue of seeing this with our less than perfect eyes and less than perfect displays.

I will use one photo as a base and you can retrieve the full size 8MP jpeg by clicking on this base image.
http://www.iisc.com/dgrin/IH9T2111_s.jpg (http://www.iisc.com/dgrin/IH9T2111.JPG)

Now this one is pretty easy to see. Both the buildings and the sky have a lot of detail. Lets look at the levels and see that graphically.

http://www.iisc.com/dgrin/PSCS_levels.jpg

Look at the intensity scale at the bottom of the level and look at the histogram of the pixels. There is a clear range of intensities at the bright end and a clear range at the dark end.

Lets see that visually by moving the pointer from the dark edge right up to the bottom of the brighter clump. That should change all of the lover values into black and make the detail in the bright section stand out.

http://www.iisc.com/dgrin/IH9T2111a.jpg


Sure enough that is what we see. So how about we go back to our original and do the same for the dark end. This time in levels grab the arrow next to the bright edge and drag it down to the edge of the clump on the dark side. What we should get is a white (blown) sky and enhanced detail in the buildings. Lets see.

http://www.iisc.com/dgrin/IH9T2111b.jpg

(See John's tutorial about combining these for extended dynamic range)

What we will do here is to split it up right in a single curves adjustment. Note you can click on the line in curves to nail down down the line as shown here. Note: I will sacrifice trueness of color to keep the example in the one curves adjustment for simplicity sake.

http://www.iisc.com/dgrin/PSCS_curves.jpg

The result from that is the following:

http://www.iisc.com/dgrin/IH9T2111_crvs.jpg

Lets go back to the program 'xv', a tool that was doing 15 years ago what so much of modern image handling is doing today. XV's equivalent to curves is even today more powerful in some ways. We will split up the intensity ranges in a more linear fashion.

http://www.iisc.com/dgrin/xv_color_editor.jpg

The result of that mod is the following image:

http://www.iisc.com/dgrin/IH9T2111_xv.jpg


In the xv intensity map it is perhaps easier to see that we have set up 2 different remapping of the intensity ranges.

To Be Continued

[mercphoto]

Very nice explanation. Thanks.

I noticed the curve you ended up with does not go through neutral gray any longer (the point 127,127). If you did so, would that have any positive impact to color accuracy? Or is that not what gave rise to the color issue you eluded to?

BTW, I like your final image better than the one you posted from xv. I don't like the color of the sky from xv. That, or my linux monitor at work is really way off.

[rutt]

The link to the original is broken for me.

[rutt]

But what I was really interested in is this: what if you wrote a very similar curve for the L channel of LAB? I think L corresponds much more closely to xv's intensity than the composite RGB curve does. (But I'm not really sure of this. I understand the L curve but not the composite RGB curve, so I don't use it.)

[cmr164]

The link to the original is broken for me.
Link is fixed.... Answers to all other questions delayed until later. I am headed out.

[wxwax]

Good stuff!

[cletus]

Great thread Charles!

Question - Any reason why did not use the pencil mode for your PS curve? I would think that it would let you make a curve with two nice linear sections, with a minimal transition between the two.

[rutt]

Expanding and overlapping the ranges of shadows and highlights in the same image is exactly what the PS/CS shadow/highlight image adjustment is great at. I used the following shadow/highlight settings:

http://rutt.smugmug.com/photos/15226436-M.jpg

and got this:

http://rutt.smugmug.com/photos/15226435-L.jpg

This image fortunately has all the detail you want in both shadow and highlight. As Charles points out, the issue is making that detail visible, increasing the contrast over the areas of interest.

Any of these things, shadow/highlight, curves, or the xv intensity curve should be seen as only the initial move in enhancing this image. The problem that all the corrections have so far is that they overlapp the dynamic range of the sky and building (this was the goal, after all.) But it decreases the contrast between the sky and building, leaving the result flat and without "pop". We'd like to fix this, wouldn't we?

One way to do that is to find a channel in which there is contrast between the building and the sky and make the curve between them steeper. The sky is blue, but the building is more blue since it's blue tinted windows are reflecting the blue sky. So I steepened the LAB B curve to make both the sky and building bluer, but the curve through this area actually increases the blueness of the building much more than the blueness of the sky:

http://rutt.smugmug.com/photos/15227094-S.gif

http://rutt.smugmug.com/photos/15227162-L.jpg

This particular image is relatively easy (I think that's why Charles chose it.) This sort of thing is much harder to get right in the presence of fleshtones, for example.

[lynnma]

Expanding and overlapping the ranges of shadows and highlights in the same image is exactly what the PS/CS shadow/highlight image adjustment is great at. I used the following shadow/highlight settings:

http://rutt.smugmug.com/photos/15226436-M.jpg

and got this:

http://rutt.smugmug.com/photos/15226435-L.jpg

This image fortunately has all the detail you want in both shadow and highlight. As Charles points out, the issue is making that detail visible, increasing the contrast over the areas of interest.

Any of these things, shadow/highlight, curves, or the xv intensity curve should be seen as only the initial move in enhancing this image. The problem that all the corrections have so far is that they overlapp the dynamic range of the sky and building (this was the goal, after all.) But it decreases the contrast between the sky and building, leaving the result flat and without "pop". We'd like to fix this, wouldn't we?

One way to do that is to find a channel in which there is contrast between the building and the sky and make the curve between them steeper. The sky is blue, but the building is more blue since it's blue tinted windows are reflecting the blue sky. So I steepened the LAB B curve to make both the sky and building bluer, but the curve through this area actually increases the blueness of the building much more than the blueness of the sky:

http://rutt.smugmug.com/photos/15227094-S.gif

http://rutt.smugmug.com/photos/15227162-L.jpg

This particular image is relatively easy (I think that's why Charles chose it.) This sort of thing is much harder to get right in the presence of fleshtones, for example.How long have I been using Photoshop?? 4 or 5 years?? and I never realized I could expand shadow/highlights options????:doh :rolleyes thanks Rutt and Charles.. good thread..