It is a beautiful evening and the light is just perfect for taking some close-up photos of the flowers in your garden. You take the time to set the exposure, frame the image, maybe even set up a tripod and a cable release too. You constantly check your histogram and keep the histogram pushed to the right, but not clipped. You’ve done it perfectly. You go back to your computer feeling very proud that you have once again captured the perfect floral photo.
It is only after you have spent some time in frustration in post processing that you begin to realize that there is a fundamental flaw with your image’s reds. They are over saturated while the rest of the image is just fine. What gives?
It is time to take a closer look at the histogram. If you look at the image above, on the left panel there are two histograms. The top one shows the histogram for the Red, Green and Blue colour in the image (we will come back to that later). Lower down you can see a second histogram in the Level’s adjustment block. This is the core of our problem.
The level’s histogram in the above diagram is adjusted to view the luminance of the image. In 99% of my photos this is a great diagnostic tool to adjust image exposure. As you can see from the total RGB histogram at the top, however, luminance does not tell the entire tale, especially with reds, it seems. (this RGB summary histogram at the top is a nice new feature on the latest release of Aperture 3)
The problem arises when we rely too heavily on the luminance histogram in our cameras. Looking at the image histogram in my camera view finder showed me that the histogram was much like that in the level’s panel. It looked healthy, but the reality was that the Red channel is clipped. This all comes down to how the luminance histogram is computed. It is compounded by the fact that on many cameras (this image was made by my Nikon D40x) only the luminance histogram is viewable. (Not a problem on the D300, which I will talk about later)
This image is a second example that was fixable. The image was in flat light and deliberately underexposed. In the above image you can see that the Level’s tool is now displaying the RGB histogram too. The grey histogram overlain on it is the luminance histogram. As this image was underexposed (and shot in RAW) there is lots of room on the right side of the histogram and we can use the levels adjustment to improve the overall exposure. Note that the levels adjuster has been moved to the edge of the red histogram, not the luminance histogram. This way we have preserved the reds in the image without clipping.
Underexposing by too much, however, makes for muddy images and more noise than we need. How do we combat this at the time of exposure and push our histogram to the right as much as possible?
On the Nikon D3oo and other high end cameras, there are more options for displaying the images in the playback viewfinder than exist on my D40x. In particular, in addition to viewing image histograms (luminance), the RGB histograms can also be viewed by changing a setting in the display function menu (check your manual). When shooting images with saturated colours (especially reds), use the RGB histogram to set the exposure and not the luminance histogram. Then you can be sure that there are no clipped colour channels.
I have talked to some very experienced photographers about this and it seems to be a common problem with the red side of the spectrum. I do not know exactly why, but I may end up doing a bit of geek research to find out. I suspect it is related to the IR filter on the photosensor and the tuned, associated red sensitivity. Fortunately, a bit of extra care using the RGB histogram in camera can help. If you don’t have RGB review in your camera, bracket your exposures and work on the image in post processing.
Comments are always welcome. Have you had problems with saturated red channels or other colour channels? How did you deal with it?