Defect, post-processing error, or just difficult light?

[blakley]

Advertisement (gone after registration)

I shoot a lot in really dark conditions with weird lighting. In particular I shoot in a lot of bars with dim incandescent light.

 

I often get good results, but sometimes I get photos which, even when color-balanced, exhibit a color cast which is not uniform but is instead "blotchy".

 

Here's an example. I shot the first photo at a local bar; you'll notice that the skin tones on the subject's upper arm are "sort of OK (maybe a little red but not too bad). But the tones on her face and forearm (circled in red in the second version of the photo) are very yellow.

 

Looking at "actual pixels" (the third image) reveals that this yellow cast is distributed in a blotchy way over the image and does not seem to be the result of uniform illumination.

 

So the question is:

 

1. Is there something wrong with my M8?

 

2. Or, is Lightroom not post-processing the image correctly (i.e. have others observed this behavior with Lightroom and been able to correct it with C1 or another DNG processor?)

 

3. Or, am I using Lightroom wrong - and could I fix the problem by using it right?

 

4. Or, am I just shooting in conditions that are too dark, underexposing, and hoping for more than I should want?

 

I'll be happy to upload the original DNG if anyone is interested in looking at the original file or trying it out with another RAW processor. I'd be grateful for your thoughts.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Shootist]

# 4, Imho.

[bradreiman]

agree with Ed. I also shoot in these conditions alot and there are usually a whole range of different light sources be they neon from bar signs, incandescents, flourescents, cell phones, candles and more. I think you've done pretty well but you'll never really get a perfect white balance with such variable light sources. not bad tho I think...B

[adan]

Bob (and anyone else who wants to shoot digital under tungsten light in very low light) Try this experiment sometime.

 

It assumes you have some way of viewing the individual channels (red/green/blue) in the image at some point. Photoshop can - I don't know whether Lightroom/Aperture can do this by themselves.

 

Open the image WITHOUT correcting the white balance. Then look at the blue channel alone. In an image like Bob's original, it will likely be entirely black, except for the brightest highlights (like the window blur back right). ZERO data. Makes one realize how amazing it is that we can change the white balance at all.

 

Basically under tungsten light, even if the 'basic" exposure is roughly correct, the blue channel is being underexposed something like 4-5 stops (i,e if the base ISO is 1250, the blue is being shot at ISO 20-40,000 or so(!!)). When the color is white-balanced, basically what one is doing is "push-processing" the blue channel that many stops to get it up roughly even with the red/green - with the same increase in noise one would get with any other push of that magnitude.

 

So then white-balance the image as well as you can, and look at the blue channel again - it usually looks like some special effect from a horror movie!

 

Using "color noise reduction" tends to blur the color channels, which spreads the noise around and results in the splotchy yellow areas (yellow being places where the blue values are still weak or absent).

 

Sometimes you can improve things a bit, especially the splotchy look, by using "Channel Mixer" to mix some of the green channel, which is usually the cleanest regarding tones and noise, into the blue channel. Usually about 70%blue/30%green with Blue as the target channel. You may have to color correct again, but at least this puts a LITTLE image data into an otherwise weak or blank blue channel.

[DaveSee]

If the "IR Issue" isn't the clue, the M8 gathers more info than one would want. Of the many pics you've posted, most are "available darkness" tending, so I'd agree with what's been posted about pushing it... HOWEVER, you've definitely got a sense of the near dark... so I'd toss out: "What ICC profiles are you using?" From the looks of it, Adobe98 or ProPhoto... the anti-reds. Pull the ECIv2 profiles...

 

rgds,

Dave

[adan]

Just to show what a RAW converter, or even the in-camera processing, must deal with in balancing tungsten light, here is a tungsten shot, ISO 1250, fairly "bright" setting (not a lot of deep shadows), showing what the 3 separate color channels look like using the native (i.e. daylight/neutral/uncorrected) balance - i.e settings that would make a gray card gray under sunlight.

 

Essentially the tungsten light contains almost no wavelengths that can penetrate the deep blue of the filters over the blue pixels in the Bayer checkerboard - those pixels are practically blind under tungsten light.

 

If you download this image, select the monochrome image marked "blue", and use levels or curves to try and lighten it to bring out detail to match the red and green, you'll see how amazing it is that we can get ANY decent color out of tungsten light.

[luigi bertolotti]

Advertisement (gone after registration)

Me too vote for #4 hypotesis : the difference (in the posted jpg) is not spectacular, and who knows HOW the ambient light is re-re-reflected on the skin... probably inside the environment you have mirrors, coloured brilliant surfaces on furnitures, metal surfaces... for me, in these kind of environment B&W is a winner... is very difficult you obtain a right color balancing.

[tashley]

Adan's right!

 

Try converting the shots to B&W in Lightroom, then move the colour temperature slider to the far right. See how the noise/grain disappears. Then move it to the far left, where the blue channel is doing all the work. See how nasty it gets.

 

That's a quick down and dirty for seeing how much potential a shot has in B&W and sometimes it can take something that looks like a paintball target and turn it into a smooth image!

 

Best

 

Tim

[blakley]

Andy,

 

Thanks very much for the detailed response. This being the case, do you think my results would improve if I were to stack an 80A or 80B on the UV/IR cut filter for these conditions, and take the hit in terms of speed? Or should I play photoshop tricks like throwing away the blue layer and replacing it with a duplicate of the green layer?

[blakley]

If the "IR Issue" isn't the clue, the M8 gathers more info than one would want. Of the many pics you've posted, most are "available darkness" tending, so I'd agree with what's been posted about pushing it... HOWEVER, you've definitely got a sense of the near dark... so I'd toss out: "What ICC profiles are you using?" From the looks of it, Adobe98 or ProPhoto... the anti-reds. Pull the ECIv2 profiles...

 

rgds,

Dave

 

Dave,

 

I'm not sure I'm following you here - I am in fact using ProPhoto RGB but I don't know what you mean by "the anti-reds"; what will using ECIv2 do for me?

 

Andy,

 

I've now done the channel examination you suggested and my blue channel looks even worse than the one you posted. It's clear from examining it that the only source of blue light comes from behind the model, and her face and forearms are in "blue shadow". The red and green channels look great though. I'm attaching first the blue channel and then the red channel for comparison.

 

I suppose that the best reaction to this kind of situation is simply to process to monochrome, though I will try using an 80B and see if I can still sustain a shutter speed that makes handheld shooting possible.

[adan]

Bob: Shocking, isn't it?

 

Yes, a tungsten balancing filter will help some - the stronger the better. I have one I tried with every digicam since the Digilux 2, and it does give the blue channel a fighting chance.

 

Also bear in mind that the lower the wattage and brightness of tungsten light, the lower the color temperature. 3200K is the color of 250W tungsten studio lights, and likely halogen studio/cinema lights (which is what an 80A or B is designed to correct), but standard indoor bulbs are even yellower. And in your shot, the bar lights are filtered even yellower by the frosted sconces, from the look of it - so you may have been shooting in light in the 1800K range or even lower.

 

Maybe Sean Reid or someone else who does architectural interiors have thoughts on dealing with indoors illumination and color.

[blakley]

It's probably even lower than 1800; the bulbs are 60 watts but on a rheostat, so when the picture was taken they were probobly running at about 15 watts.

[robertwright]

for moving subjects the 80B is going to kill your shutter speed in that environment.

 

you didn't say what ISO this was shot at. This might be a good time to try "push processing" in Capture One, you can shoot at 160 or 640 and allow it to be a stop or 2 stops dark. IOW, get an idea of what the exposure should be and then turn the ISO down not up. You will see dark pictures on the LCD and the histogram will be well to the left.

 

Capture One will let you open this up almost 2 stops with little penalty. 160 has an incredible dynamic range. You may get better results, especially in mixed lighting.

 

Another point is do you remember film? Blotchy shadows and grain were pretty normal. Color is rarely uniform, and the more uniform it is the more it looks unreal or plastic.

 

noctilux anyone?

[blakley]

Actually this was shot at ISO 640 with a Noctilux (50/1.2 Noctilux "ASPH"). Film results wouldn't have been better - I've tried film in the same bar and when the lights are turned down, the only option is B&W (typically Ilford Delta 3200).