Getting The Best Exposure With The Least Noise And Greatest Dynamic Range

Digital cameras in general behave differently than film cameras in their handling of highlights and shadows. With film, generally one would meter so that the midtones of the image were exposed as midtones, and let the shadows and highlights fall where they may. The response curves of film have a fairly smooth and gradual decrease in response in the shadow and highlight areas. Digital is different; the sensor response is even until it reaches its maximum output value and then the response of the sensor hits a brick wall--it just stops responding to any additional increase in brightness. Reaching this point is called clipping, and results in very ugly highlights that are either flat white with no detail or have unattractive color shifts if only one of the 3 color channels is clipped. On the other hand, an amazing amount of image detail can be extracted from shadow areas, especially if shooting RAW mode. So with digital, the best exposure strategy is to keep highlights from reaching the maximum output output value of the sensor (except for small specular highlights that do not detract from the image if they are clipped). But at the same time, you want to keep the highlights close to the clipping point; the closer the highlights are to clipping, the less sensor noise will be visible in the image. A digital sensor has a fairly constant noise level at a given ISO setting; the greater the signal strength, the higher the signal-to-noise ratio will be. At least until the signal begins to be clipped, that is. Ideal exposure with a digital camera (capturing the greatest possible dynamic range with the lowest possible level of noise and and other undesirable digital artifacts at a given ISO setting) is always that which causes the RAW data to closely approach, but not reach the clipping point for non-specular highlights. By "closely approach" I mean that the non-specular highlights are within the camera's minimum exposure setting increment (typically 1/2 or 1/3 stop) of clipping.

There are a few additional principles to consider when selecting the ISO setting for an exposure:

Using The Histogram To Properly Judge Exposure

The best tool the digital photographer has for evaluating exposure is the histogram, but there are a few issues with histogram displays that one must be aware of in order to properly interpret them to determine ideal exposure. Camera histograms are always based on the camera's JPEG conversion settings, even if you are shooting RAW. This means that there are differences between the camera histogram and the underlying RAW data that must be understood properly before ideal exposure can be consistently achieved. I recommend setting contrast to its minimum value and using the Adobe RGB color space if the camera supports it. This will help the camera histogram match the underlying RAW data as closely as possible. It will also ensure that when shooting JPEGS, they will capture the widest possible dynamic range and color gamut.

When shooting JPEG images, what you see is what you get, with one little gotcha. When the brightest part of the image is not white, you need to leave a little extra room on the right edge of the histogram if the histogram does not display each color channel separately. The standard luminance histogram display is a mixture of approximately 30% red, 59% green, and 11% blue. This means that if the brightest part of the image is a saturated red or blue, it is very easy for the red or blue channel to clip without the histogram appearing to be overexposed. If your camera displays a separate histogram for each color channel, like the Canon 1D Mark II, you don't need to worry about this, because the the color channels are not averaged together into a single display.

When shooting RAW, things get a little more interesting. There is almost always a difference between the clipping point of the camera histogram and the clipping point of the RAW data, because the camera histogram is based on the in-camera JPEG conversion of the underlying RAW data. In-camera JPEG conversions typically throw away 1-2 stops worth of the sensor's dynamic range, which means that there will typically be up to a stop of extra usable headroom in the highlights when you shoot RAW. The rest is extra shadow detail you get only in the RAW data.

The bad news is that the exposure interval between the clip point indicated by the camera histogram and the actual RAW data varies from camera to camera, depending on the camera model and its' contrast, color space, and saturation settings. The good news is that determining the exposure interval between clip points is fairly easy. All you have to do is put the camera in manual mode, set it on a tripod, and aim it at something that has some nice neutral whites and some colors. A Gretag-Macbeth Color Checker chart is an excellent choice. Set exposure so that the camera histogram has a lot of room on the right edge, indicating underexposure. Shoot RAW (RAW + JPEG mode is recommended if your camera has it) and increment exposure between frames in 1/3 stop intervals. Make a note of the first shot where the camera histogram data touches the right edge of the display. Continue shooting and increasing exposure for another two stops. Then open the RAW files and find the first frame with clipped RAW data. Look for spikes on the right side of the converted image histogram data that won't go away even if you adjust the exposure setting way down. The exposure difference between the first clipped RAW image and the first image showing clipping in the camera histogram is what you're looking for.

My 1Ds and 1D-MkII have about a 2/3 stop exposure interval between the camera histogram clip point and the RAW clip point, so when shooting RAW, I set exposure so that I have a few pixels of blinking highlights, but don't have a big spike on the right of the camera histogram. This means that my non-specular highlights are about 2/3 of a stop from being clipped. I still have a bit of wiggle room if I accidentally overexpose a bit, but I'm not underexposing so much that noise levels and the dynamic range of the capture are negatively affected. Exposing more than a stop below the RAW clip point starts noticeably degrading image quality.

It's usually not a crime to let a white light clip to pure white if the rest of the frame is correctly exposed as a result. Colored lights can be more problematic; you may see color shifts near the light where one color channel clips before another. For example, a red light shining on a wall may display a red area away from the light, with a yellow fringe between the red area and the pure white clipped area of the light proper. You'll have to exercise some creative judgment to weigh the negative effects of the color clipping from overexposing the light vs. the increased noise level from underexposing the rest of the scene. You will encounter this dilemma to a lesser extent when shooting things with extremely saturated colors, such as macro shots of flowers. You will often have to underexpose the shot a little more than otherwise to keep from clipping any of the color channels, but you don't want to underexpose too much or you'll increase sensor noise to potentially objectionable levels. If you must shoot JPEG, use the Adobe RGB color space setting (color matrix 4 on the 1Ds) to minimize this as much as possible. Shooting RAW will give you the maximum flexibility in dealing with this dilemma; RAW offers 100% of the gamut of the sensor and the greatest ability to extract detail cleanly from the shadows.

Metering Modes, Camera Modes, And The Histogram

Once you understand exactly what the histogram is telling you, determining correct exposure is fairly straightforward. Don't get too wrapped up in what the camera's meter or a handheld meter indicates should be correct exposure. The ultimate indication of correct exposure is the histogram, with the caveats outlined above. Depending on the subject reflectance and the camera meter's characteristics, ideal exposure may be more than a stop above or below what the meter says. I recommend using the camera meter as a guide to get initial exposure settings within the realm of common sense, and then fine-tune settings based on the histogram. Shoot a few "digital Polaroids" (they're not costing you like they would with film) and adjust as necessary.

In situations where the lighting is constant or the primary light source is flash, I recommend manual mode and spot metering on the brightest subject you don't want blown out. Adjust shutter speed, aperture, and flash exposure compensation to get as close as possible to ideal exposure without clipping. For concerts, partly cloudy outdoor shots, or other situations where the lighting can vary significantly from shot to shot, I recommend aperture priority or shutter priority mode (I prefer aperture priority) with either evaluative/matrix or center-weighted average metering, depending on the subject. Adjust exposure compensation and flash exposure compensation (if using flash fill) to dial in exposure and the amount of flash fill desired. But however you shoot, a properly-interpreted histogram is your best guide to determining ideal digital exposure.

Related articles: