Color correcting JPGs with major white balance problems: Can the results ever be any good?

[tom_mann1]

<p>In the thread, http://www.photo.net/beginner-photography-questions-forum/00Zh6L, the OP submitted an image, http://static.photo.net/attachments/bboard/00Z/00Zh6L-421643584.jpg, which was taken under strongly red gelled lighting. Several of us tried to adjust the color to get reasonably natural skin tones. I think it's safe to say that none of the attempts, mine included, were very successful. That led to a discussion of the possible reasons for the difficulty. Tim L proposed that a major part of the difficulty occurred because the lighting was narrow band (having been filtered through a gel).<br /><br />I'm in complete agreement with Tim that this is an important factor. However, I think that the problem is more broad than this. Specifically, I think that it is surprisingly difficult to do any major color correction work on JPGs, even if the lighting is broadband, just with a substantially different color temperature than that set in the camera or during post processing in the raw converter. <br /><br />Attached to this post is the in-camera JPG version of a test shot of a dish strainer (containing a nice assortment of various colored objects) using daylight WB in a d700. The lighting was from typical 60 and 100 watt household incandescent bulbs. As expected, it has the usual orange color cast. <br /><br />I saved both JPG and NEF files for this image. In the next post, I show what the image looks like when I run the NEF for this image through ACR, and simply put the eyedropper for ACR's automatic color correction on a white part of the strainer. Other than sharpening after resizing, and converting to sRGB for web display, nothing else was done to this image. As you can see, in one simple step, the colors have become very reasonable / realistic. <br /><br />I would like to see how well the experts here on photo.net can do with the orange colored JPG. If you obtain good results, I would love to know how you did it. However, if no one comes up with decent results starting with the JPG, the results of this little exercise speak not only to processing our own photos, but to the importance of using raw files, color correction in mixed lighting, and responding to folks who post JPG images on photo.net asking for help with major color problems. <br /><br />Have fun!<br /><br /><br />Tom M</p><div></div>

[tom_mann1]

<p>.</p><div></div>

[Matt Laur]

<p>Wow, Tom, that JPG is toast. There's just not enough <em>there</em> there to keep the red-ish stuff (like that mug) there while getting those whites out of the headed-towards-IR-area they're in. <br /><br />I presumed you wanted to avoid a masking nightmare, which leaves us with various ways to torture those 8 bits into something more neutral across the board. My bag of tricks isn't letting me get there without throwing a lot of things into blue-ville. Without masking, I can't think of an elegant way to even approach what you can do with the latitude of the NEF file.</p>

[Matt Laur]

<p>Wow, Tom, that JPG is toast. There's just not enough <em>there</em> there to keep the red-ish stuff (like that mug) there while getting those whites out of the headed-towards-IR-area they're in. <br /><br />I presumed you wanted to avoid a masking nightmare, which leaves us with various ways to torture those 8 bits into something more neutral across the board. My bag of tricks isn't letting me get there without throwing a lot of things into blue-ville. Without masking, I can't think of an elegant way to even approach what you can do with the latitude of the NEF file.<br /><br />Of course, that would make <em>fine</em> black and white! Problem solved! :-)</p>

[Glenn McCreery]

<p>Here is what I get in a couple of minutes using hue-saturation and brightness/contrast in Photoshop. Blue and green saturation was increased and red and magenta were decreased. Contrast and brightness were both increased. Blue and green colors are out of whack compared with your NEF image. Someone more knowledgeable can probably do better with these colors.</p><div></div>

[tom_mann1]

<p>Matt, the problem is exacerbated because the color temp of household tungsten lights are so low. However, I used them, not something with a higher color temp because household tungstens are a common source of WB problems. </p>

<p>The success of the RAW conversion in achieving a decent WB really drives home how important it can be to shoot raw when you don't have control of the lighting.</p>

<p>B&W conversion ---> ;-)</p>

<p>Tom</p>

[tom_mann1]

<p>Glenn, thanks loads! I tried several methods and all of them resulted in images similar in color quality to yours. I didn't want to post my results of tweaking the JPG till at least a few folks weighed in with their methods. </p>

<p>Thanks again,</p>

<p>Tom M</p>

[Tim_Lookingbill]

<blockquote>

<p>I'm in complete agreement with Tim that this is an important factor. However, I think that the problem is more broad than this. Specifically, I think that it is surprisingly difficult to do any major color correction work on JPGs, even if the lighting is broadband, just with a substantially different color temperature than that set in the camera or during post processing in the raw converter. </p>

</blockquote>

<p>If it was a JPG issue as opposed to working on the Raw version, then how do you explain the green WB color correction of the rock climbing wall image performed on the posted jpeg shown in this thread...</p>

<p>http://www.photo.net/beginner-photography-questions-forum/00YmG4</p>

<p>I still don't think you could fix the red light image (or make it look like it was lit by white light) whether on the JPEG or Raw. Scott already showed all he could do on his drummer shot but he didn't indicate if he worked on the jpeg or Raw.</p>

<p>BTW I couldn't make your posted tungsten image look like your corrected version in ACR. I still can't remember how I went about fixing the green rock climbing wall jpeg in ACR. That image seemed more pliable than your posted tungsten version.</p>

[tom_mann1]

<p>Hi Tim -</p>

<p>Thank you for joining this discussion. To address the point you raised, I think that the difficulty in cc'ing problematic images can come from several different root causes, and they can all be operating at the same time in various proportions:</p>

<p>(a) trying to make big WB changes to 8 bpc JPGs,</p>

<p>(b) trying to introduce hue and saturation variation where there was none because of the narrow band spectrum of the light; and,</p>

<p>© probably several other mechanisms that we haven't yet identified and discussed.</p>

<p>So, in the case of the red gelled light image, we were fighting both problems (a) AND (b).</p>

<p>IMO, the most likely reason that you had more success correcting the green rock climbing wall is almost certainly that high pressure metal halide lamps were being used, and while the spectrum of these is far from smooth, there is significant output across the visible spectrum. There are nice representations of the spectrum of such lamps in the Wikipedia article on them: http://en.wikipedia.org/wiki/Metal-halide_lamp. If they had been using only low pressure Na or Hg lamps in that gym, I will bet you a steak dinner that you wouldn't have had as much success. :-)</p>

<p>The extended but bumpy spectrum of high pressure metal halides is in contrast to the household tungsten lamps that lit my test shot. The tungsten lights, of course, have a continuous and beautifully smooth spectrum, but, in contrast to the metal halide lamps, because their color temp is so low, household tungsten lamps have almost zero output in the blue part of the spectrum, so, there is just nothing to work with, or, as you put it, the image illuminated by them is less pliable.</p>

<p>In a few minutes, I am going to start another digital darkroom thread that deals with problem (b), above - narrow band illumination. I'll post the link here as soon as I am finished writing it.</p>

<p>Cheers,</p>

<p>Tom M</p>

[tom_mann1]

<p>The thread on the effects of no hue and saturation variability is here:<br>

http://www.photo.net/digital-darkroom-forum/00ZjNZ</p>

<p>Tom M</p>

[sarah_fox]

<p>I can't do any better than Glenn. If you look at the individual curves, blue peaks and dies out at relatively low levels, and green at moderate levels. The red curve is truncated to the right (clipped). If the exposure were knocked down by a stop or two, I might have more success adjusting the curves. However, as it stands, the red is always going to be out of whack with the other two channels, being softly clipped in the jpg.</p>

[tom_mann1]

<p>Sarah, you hit the nail on the head. If the only form of image one has to work with is a JPG, it often will be clipped on either or both ends of the histogram. I'm preaching to the choir, but, of course, once this happens, just like in my dish strainer test shot, you can never recover the information contained in the levels that were smooshed together. </p>

<p>To illustrate this, I took the NEF for my dish strainer test shot into ACR, dropped the exposure a bit, added a fair amount of "recovery" and "fill", and, most importantly, set the black point to zero instead of its default value of 5. I didn't touch the WB in ACR - I let the WB remain: "As Shot".</p>

<p>The result is the very low contrast JPG image attached to this post.</p>

<p> </p><div></div>

[tom_mann1]

<p>The histogram for this new JPG confirms no clipping...</p><div></div>

[tom_mann1]

<p>Now, if you bring *this* JPG into PS, unlike the in-camera JPG, you can use simple, automatic tools (eg, the eyedroppers in "curves" or "levels") to do a very fast and credible job of white balancing the JPG, an almost impossible task with the in-camera version.</p>

<p> </p><div></div>

[tom_mann1]

<p>In fact, in the specific case of folks who submit images to photo.net asking for help, we are almost always stuck with trying to deal with only a poorly exposed JPG because photo.net's posting mechanism supports only JPGs in threads and doesn't provide as obvious and straightforward a mechanism to transfer RAW files.</p>

<p>For me, one of the major results of this exercise is the realization of how difficult it can be just to WB a JPG, and how relatively minor changes in the look of the JPG can change the PP processing problem for JPGs from essentially impossible to trivially easy.</p>

<p>The above is known to many photographers who often recommend shooting very flat in-camera images, keeping the doors open for later adjustments, but, IMHO, this conclusion should be emphasized with less experienced folks who, for some reason, don't see the advantages to saving RAW files as well as JPGs.</p>

<p>Cheers,</p>

<p>Tom M</p>

<p>PS - After I posted the above, I was glancing through other threads and was very happy to see the one titled, "Learned the hard way to shoot RAW" that is discussing some of the same issues. :-)</p>

[nathansanborn]

<p>Fixed! That was easier than I thought!</p><div></div>

[tom_mann1]

<p>:-)</p>

<p>Just don't forget to send that old painter over here with a few buckets of colored paint when he is done with the Sistine Chapel. ;-)</p>

<p>Tom M</p>

[JohnWebster]

<p>I followed the first link to this red image---it is sort of correctable----lots of futzing with curves, levels, saturation----and pushing the rgb numbers around to give me something close to a caucasioan skin tone.</p><div></div>

[tom_mann1]

<p>Hi John - I think you may have inadvertently put your last post in the wrong thread. That image was being discussed here: http://www.photo.net/beginner-photography-questions-forum/00Zh6L .</p>

<p>Cheers,</p>

<p>Tom M</p>