Removing specular highlights in CS3:

[Dinarius]

I have a problem with a shot of an oil painting.

Where the surface of the canvas is uneven (it's an old painting) there is a shine (specular highlights) from the varnish that was applied over the paint.

How can I remove this?

Many thanks.

D.

[mikeguil]

Almost impossible without harming the integrity of the painting.  It would be best if you could reshoot it and position your lights so that you don't get the 'shine'.  Use  soft, diffused light at no less than 45 degree angle to the painting.  You could also polarize the lightsource with sheets of polarizing material - but that can get expensive.  I've bounced the lights of the side walls of a small studio (rather than using softbox or umbrellas) and achieved good results too.

[Jonathan Wienke]

Forget about doing it in Photoshop; the best solution by far is to light the artwork properly in the first place. A pair of softboxes covered with polarizing gels positioned at about a 40-degree angle to the painting (closer to parallell than perpendicular) and a polarizing filter on the camera lens is your best bet. Set up the lights first, then adjust the polarizer on the camera for minimum reflection, then place the polarizer gels on the lighting at a 90-degree angle to the camera polarizer. It may sound like a hassle, but it's far easier than manually repainting every specular glare in a realistic-looking manner after the fact.

[framah]

Yeah, what Johnathon said as that is how the pros do it. If you want it right do it right.

[George Machen]

If a reshoot isn't possible, take a look at Dan Margulis' Lab book, pp. 170-174, or this on-line rehash:

Chapter 8: The Impossible Retouch; Fixing blown skies, facial highlights and more
http://www.dgrin.com/showthread.php?t=21648

The technique takes advantage of Photoshop's compromise in dealing with an "imaginary color" when converting from Lab to RGB: In any specular or blown-out highlight in RGB, the white area inherently *must* be neutral, R=G=B. But in Lab when that area's luminosity L is 100, it nevertheless still can have non-neutral, even fully-saturated color values in the color ab channels. Consequently, Photoshop reconciles the increasing impossibility of a non-neutral highlight as it approaches a specular by "splitting the difference" when converting to RGB. The crux of the procedure, allowing the Color blending mode to merge (or "Flatten Layers") when doing the Lab>RGB conversion, harnesses this effect to nicely fix blown-out highlights, whereas the same blending mode perforce would have to fail if done in RGB.

[Dinarius]

Thanks for the tip about using Lab. The amount of glare is minimal and we can probably live with it.

For the record, no designer or artist I've ever worked with likes oil paintings shot using polarizing gels when compared side by side with a tranny or file which didn't use them. They ruin a painting, massively increasing saturation and flattening the surface to the point where it looks like a print. Almost all surface texture, which is what oil painting is about, is eliminated. Contrary to common belief, galleries and museums only use polarizing gels when they absolutely have to.

D.

[bdawg07]

I completely disagree with you. We shoot hundreds of paintings a year and 9 times out of 10 they are shot with polarizing gels. The gels do not ruin a painting by over saturating the colors or flattening the surface texture. We shoot for galleries and museums on a regular basis and use the gels all of the time.

Your results are largely dependent upon:
1) Experience in shooting art
2) Lighting techniques
3) Camera/Capture system
4) Accuracy of capture & output profiles

With that said, we will occasionally shoot without the gels when a painting contains metallics or subtle details in the shadows.


Adam Brown
Osio-Brown Editions
http://osiobrown.com/giclee_printing.html

Thanks for the tip about using Lab. The amount of glare is minimal and we can probably live with it.

For the record, no designer or artist I've ever worked with likes oil paintings shot using polarizing gels when compared side by side with a tranny or file which didn't use them. They ruin a painting, massively increasing saturation and flattening the surface to the point where it looks like a print. Almost all surface texture, which is what oil painting is about, is eliminated. Contrary to common belief, galleries and museums only use polarizing gels when they absolutely have to.

D.
[a href=\"index.php?act=findpost&pid=137511\"][{POST_SNAPBACK}][/a]

[Dinarius]

Different work practices on each side of the pond, I guess! ;-)

But, was it ever thus!?

D.

[Jonathan Wienke]

For the record, no designer or artist I've ever worked with likes oil paintings shot using polarizing gels when compared side by side with a tranny or file which didn't use them. They ruin a painting, massively increasing saturation and flattening the surface to the point where it looks like a print. Almost all surface texture, which is what oil painting is about, is eliminated. Contrary to common belief, galleries and museums only use polarizing gels when they absolutely have to.

Very atypical, and not very correct, either. The only reason saturation goes up is because you are eliminating veiling glare, same as shooting something under water from above the water surface. If you use a polarizer to eliminate the water surface glare, you'll get color a lot truer to the subject than if you don't remove the surface glare. As to "flatness" and loss of texture, that's a lighting issue and NOT a polarization issue. If you're not picking up as much surface texture as you'd like, don't diffuse your lighting as much. Use smaller softboxes, or a simple small round reflector on the strobe head instead of a softbox. And don't have both lights equal intensity, have one light a stop or two brighter than the other one so you get a main/fill effect like you do in portrait lighting. You'll have to back the lights further away from the painting to keep the overall illumination even, but that's another trick to pick up more surface texture detail. The bottom line is that getting things right in-camera with the right lighting is far better than trying to fix in Photoshop after the fact.

[Michael Bailey]

Dinarius,

Regarding the polarizing issue, I'll throw in my two cents. I think both camps are correct, to a degree. Perfect polarization does indeed eliminate glare spots and does increase apparent saturation and contrast, so your photo may be accurate to the technical composition of the original's color and density, but maybe not to the way the painting is perceived.

In the past I had a painter client who used a vicious amount of texture and varnish. Double polarizing raised the contrast so high that the film literally could not maintain detail in both ends of the tonal scale. I had to resort to "flashing" the film--exposing it to a plain white board at a 3-4 stop underexposure--to bring the contrast back under control. Now, whenever someone calls me about photographing their art, my first question is not, "How big is it?" or, "Is it in a frame?" No, the first thing I ask is, "What's the surface like?"

Digital photography has made this sort of thing easier, but I'd still rather stay away from the world of impasto.

As for a possible, simpler, solution to the glare problem you already have, perhaps you could rely on the fact that your hot spots are probably the only part of the file that measure 255-255-255. Use that fact to make selections and then...well, then I'm not sure. Maybe...

-increase the selected areas by a few pixels, and feather them
-invert the selection
-Ctrl (or Cmnd)-J to put the selection on a new layer
-at this point you'll have a Background layer, and a layer above which has holes in it where the hotspots happened
-on the Background layer (or, more safely, a copy of it) apply a Median filter (Filter>>Noise>>Median). I think that's the one you'd want because it would blur the layer in such a way as to push the darker, good, values into the light, hot, areas.
-The darkened values would appear under the holes in your top layer.
-With some experimentation as too what numbers to apply to the different steps, I think you could fill in the holes pretty well.

For what it's worth.

MB

[Morgan_Moore]

I have a problem with a shot of an oil painting.

Where the surface of the canvas is uneven (it's an old painting) there is a shine (specular highlights) from the varnish that was applied over the paint.

How can I remove this?

Many thanks.

D.
[a href=\"index.php?act=findpost&pid=137409\"][{POST_SNAPBACK}][/a]

Here is a bodgers method for sorting out your file..

Create a new layer from background in photoshop

'replace colour'  'click on a  highlight/reflection' replace it with a darker colour

Use eraser tool to delete that layer in the unaffected areas so the actual highlights (light parts of the image) are not affected

flatten image

---

You could also create new layer, heavily gausian blur the background layer, select the highlights (in the top layer) wiht the 'magic wand',  ' selection feather' to soften the edges of you selection and hit delete


---

Giving a quick try you can select a 'colour range' click on a highlight this with then pick up those highlights, feather the selection and delete that selection with a heavily blurred layer behind

----

There must be other 'filtration' methods of grabbing the highlights and deleting or replacing them with the content of another layer like a blurred one


I am interested to know those if any one can chip in

S

[Dinarius]

If nothing else, this has thrown up a few interesting PS techniques. ;-)

Another workaround that looks good to me.........

1. Create background copy.

2. Enlarged the affected area to pixel visible levels.

3. Choose Select/Color Range.... and click on a flared/highlighted pixel. Click OK. Now all similarly affected pixels will become active.

4. Open a Levels Adjustment Layer. Click OK.

5. Change the mode to Luminosity. Reopen the Levels layer.

6. Now adjust to taste. This will usually involve moving the shadow slider in to the toe of the histogram and darking the mid-tone slider a little.

On the painting I had to work on this worked extremely well. I was left with the clear feeling/look that this was an oil painting. It did not look flat, and the offensive flare had been all but eliminated.

If anyone has any improvements to the above technique, please post.

D.

Isn't the Color Range tool one seriously powerful piece of kit?! ;-)

[Digiteyesed]

Forget about doing it in Photoshop; the best solution by far is to light the artwork properly in the first place.

Depending on whether the image was shot in RAW and the severity of the blowouts, they might be recoverable. For example, here's what I recovered from one image with blown out areas using Photoshop and Photomatix:

http://www.urbanrefugee.ca/example08.php

[Wayne Fox]

My experience in copying paintings is limited to a couple of friends that ask me to make prints of their personal work, which I did by using direct sunlight.  I assume there are some very good techniques since many artists make litho's of their work to sell.

The idea that crossed my mind is probably physically impossible because of the accuracy required.  But hey ... I don't know what I don't know, so I thought I'd throw it out there.

Basically it's just a version of an old trick to get rid of specular highlights when scanning a photograph on some type of E or Linen paper.  Most scanners today don't have as much problem, but back 10 years ago, you had a hard time getting a good scan - usually covered with 100's of white spots - specular highlights from the prints texture.  Some photographers even ran their proofs through very heavy texture machines to make it even harder to scan.

Many of you probably know about this, basically you scanned the image, then turned it 180 degrees and scanned it again.  After aligning the two images in Photoshop layers, you change the blend mode of the upper layer to darken ( I think it was darken).  Anything white in that layer lets what is underneath show through. Since the specular highlights rarely occurred in the same place, the effect was a spot free print.

The challenge  doing this with a camera and a painting would be to have enough accuracy that the 2 images would align perfectly. I assume it would be easier to rotate the camera, but all kinds of things would probably crop up, including distortion, vignetting, etc.  

As I said, I don't know what I don't know

[PeterLange]

If a reshoot isn't possible, take a look at Dan Margulis' Lab book, pp. 170-174, or this on-line rehash:

Chapter 8: The Impossible Retouch; Fixing blown skies, facial highlights and more
http://www.dgrin.com/showthread.php?t=21648

The technique takes advantage of Photoshop's compromise in dealing with an "imaginary color" when converting from Lab to RGB: In any specular or blown-out highlight in RGB, the white area inherently *must* be neutral, R=G=B. But in Lab when that area's luminosity L is 100, it nevertheless still can have non-neutral, even fully-saturated color values in the color ab channels. Consequently, Photoshop reconciles the increasing impossibility of a non-neutral highlight as it approaches a specular by "splitting the difference" when converting to RGB. The crux of the procedure, allowing the Color blending mode to merge (or "Flatten Layers") when doing the Lab>RGB conversion, harnesses this effect to nicely fix blown-out highlights, whereas the same blending mode perforce would have to fail if done in RGB.
[{POST_SNAPBACK}][/a]

Nice technique, for example to recover burned skin tones.
It can be favorably combined with the S/H tool to recover some lightness differences as well.

Said “imaginary colors” due to integer Lab encoding are illustrated here:
[a href=\"http://www.brucelindbloom.com/LabGamutDisplayHelp.html]http://www.brucelindbloom.com/LabGamutDisplayHelp.html[/url]

Peter

--