RAW Dialog: Hilite Recovery problem

[MarkT]

Hi Andreas,

I can only speak for Nikon NEF/NRW files, but here is the same image with 100% "Highlight Protection" applied in Nikon Capture NX2:

NX2_highlight_protection-1.JPG (22.29 KiB) Viewed 6986 times

UFRaw 0.16 does not have a separate highlight recovery/protection function (that I can find...), but with simple exposure compensation applied the highlights can be restored without colouration:

UFRaw_exp_comp-1.JPG (20.07 KiB) Viewed 6989 times

If I am missing some information I look forward to hearing about it.

Mark

[mjdl]

Just some rapid reaction: these highlight recovery casts have been discussed before: e.g. in http://www.dl-c.com/board/viewtopic.php ... &sk=t&sd=a where the ever resourceful Den recommends a couple of approaches to fix it after the RAW has been processed.

I recall fiddling around a few months ago with the Olympus RAW image published on Wrotniak's Olympus enthusiast website at http://www.wrotniak.net/photo/43/e30-hirec.html. I used both PWP (5.0 release) and RawTherapee.

With PWP, any colour model, and any positive amount of highlight recovery, the clapboard wall on the right of the photograph has a screaming yellow cast. With RT, using Luminance highlight recovery, no cast is introduced, CIE-Lab introduces a barely perceptible amount of magenta, while the Colour Propagation highlight recovery process does produce a small visible magenta cast, not surprising if there is a large area of clipped pixels so that there are really no nearby unclipped pixels that can be propagated (if I've understood the process correctly...)

I drew the hasty conclusion then that PWP was using some kind of colour propagation recovery method, although I've no real serious knowledge of the different procedures it does use.

Conceivably there might also be contributions from colour space conversions of clipped pixels: each colour conversion engine has its own way of handling extreme cases.

[den]

Some thoughts...

Too often when editing "problematic" images, I believe that there is a tendency to rely too much on a single PWP transform, a single PWP RAW conversion, or with PS finding an appropriate ‘action(s)’ rather than analyzing what is to be done to optimize the available image data for a given digitally acquired photo... and the need for compromise in order to blend/balance the visual elements for an overall satisfactory editing outcome.

The OP has noted that in the case of Highlight recovery, while more highlight detail is achieved, the shadows and mid-tones have been compressed… a necessary compromise.

Another poster has commented that for Canon CR2 PWP RAW conversions with Highlight Recovery, there is a tendency for magenta/cyan hues in the recovered image areas… necessary assumptions for otherwise non-recorded data.

I have found that PWP5’s RAW conversion Highlight Recovery feature to provide more than adequate additional detail in otherwise clipped image areas for my Canon 350D CR2 RAW files when:
(1) Two RAW conversions are made. One for Highlight Recovery and one for normal conversion of the shadows/mid-tones. These two conversion files are blended in StackImages or Composite-Blend using symmetrical and/or asymmetrical tone range masks;
(2) Magenta/cyan hues if present are removed using the Paint Tool in HSV Hue Only mode with an appropriately selected hue;
(3) Final adjustment of brightness, contrast, and color to preferences; and
(4) “Creative” Local Contrast Enhancement and Sharpening.

The settings/masks/Paint Tool for the above process are image specific and not necessarily generic for other images.

I consider PWP to be invaluable, providing the access to image data and the tools to edit to specific needs… and not a “problem” when no single edit solution is possible to optimize a “problematic” image.

Illustrated below: Left – a Canon 350D camera JPEG image; Right – its CR2 RAW file processed as described above…

Please ask, if there are questions… although a message board is a poor venue to share detailed techniques/workflow… the illustration is meant to show one possible PWP solution for the image and I would be more than willing to use another’s rather than one of my own.

[keithrj]

I have been watching this thread with interest but when I went to revisit my original problem image I got a completely different result to what I got when I started this thread. My original image changed so dramatically with any of the hilite recovery settings (except none) that I decided to post a question. Originally, the image darkened a good 2 - 3 stops at all settings but now all settings actually lighten the image slightly.

Very strange happenings - so I will continue to experiment to see if I can find out what is going on.

One thought, if you do not set hilight recovery and generate a 48 bit tif is there still all the information in the tif image to do hilight recovery using some other method? As my RAW images are 14 bits per channel (42 bits total) my first guess is that all the information is retained in the generated 48 bit tif.

[keithrj]

OK, I think I have nailed it down to the particular photo I am using. Firstly I appologise for my previous post - I erroneously used a different image to my original one. I have now revisited my original image and although I am getting the same result as my original post I do see that the highlights have been recovered but in doing so the whole image has darkened considerably as indicated by the histogram in the attached clips. By increasing the midtones I can recover the image back to a pretty good level but with the highlights fully recovered.

Please note that the setting in the clips are my Camera Default to get an image that is very close to the camera JPG.

No hilite

Normal.jpg (20.59 KiB) Viewed 6771 times

Level 1

Level1.jpg (18.4 KiB) Viewed 6783 times

[den]

Perhaps an alternative to image blending the 'Normal' and 'Level1' image conversion versions posted above in order to recover to preferences the 0 to 55% toned shadows/mid-tones is to use a ColorCurves transformation with a brightening contrast curve and a 0->55% asymmetrical tone range mask on the 'Level1' conversion image

Level1_shdw-mt-recvry_400px.jpg (48.4 KiB) Viewed 6741 times

The point being to not try to do too much at one time in the RAW Dialog so that the highlight and the shadow/mid-tone tone ranges are separately adjustmented for optimum brightness, contrast, and color without compromises that may occur if attempted in the same transformation iteration.

[keithrj]

Den, your solution works well but I do try to get as close as possible to my final image in the RAW conversion process as I have a hunch that once the raw image is converted there may not be as much info in the image to work with as there is during the raw conversion process.

With that aside, what I have found is that the hilight recovery in PWP RAW conversion is much more effective than in other converters such as Lightroom which could not get anywhere near what PWP gave me. So for difficult images where it is important to recover hilights I will be turning to PWP even thought more work is required to get the overall image back to where I want it.

[Charles2]

Sometimes a quick and dirty way to deal with blown highlights is to use no highlight recovery in the Raw dialog, then while working with the TIF, color remap the extreme HSV value of those pixels to something acceptable.

Useful, for example, for small white areas like bits of sky, a wall of a house or a car in the background, etc.

[tomczak]

I'm not sure which parts of dcraw are used in PWP, but here is a discussion on possible combination of highlight modes and the sensor saturation settings in dcraw that determine how RAW channels are scaled and that can cause the magenta casts in the hightlights (section on Sensor Saturation Levels):

http://www.guillermoluijk.com/tutorial/ ... dex_en.htm

[kr@utg]

Dear all,
1. ) Would it be possible to add the saturation level of dcraw as a parameter to the raw dialog ?

-S <num> Set the saturation level

2.) Would it be possible to add the H = 2 (blend) option of dcraw as a parameter to the raw dialog (to insure that the highlights are not yellow or magenta but white) ?
-H [0-9] Highlight mode (0=clip, 1=unclip, 2=blend, 3+=rebuild)

I am using another program (http://www.fixfoto.info) allowing to set exactly these 2 parameters of dcraw for better clipped highlight color.

Best regards
Michael