Monday, October 8, 2007
Exposing For RAW
There are special considerations to take into account if you're shooting RAW and you want to be sure that you're getting a proper exposure
Shooting in controlled situations, where you can light the scene and use an external incident meter based on the exposure tests described and perhaps bracketing, may better lend itself to the ETTR technique. Shooting quickly in the field, where you have to rely on the camera metering and where highlight values may change at any given time, may require you to live with a bit more noise in the shadows.
Noise in shadows can be reduced in a number of ways, but if you blow out highlight data you wish to capture, there's nothing that will bring that data back. One way to reduce shadow noise is to increase the black clipping in a RAW converter, especially if there isn't any critical deep shadow detail you need to reproduce. Much of the noise will clip to black, effectively removing it from the image. This technique won't work if your goal is to render as much shadow detail as possible. Downsizing a high-resolution image also will reduce noise, since a group of adjoining dark pixels will be sampled into a single pixel value. If you can capture more than one shot of the scene, at the same exposure, you can align multiple layers in Photoshop CS3 as a Smart object and use the Median blend mode to remove noise (see http://photoshopnews.com/2007/03/27/image-stacks-in-photoshop-cs3-extended/). Lastly, there are third-party noise-reduction plug-ins available for use within Photoshop.
With respect to bracketing exposures when possible and desired, it would be useful if our cameras could bracket in one direction: from “normal” to increased exposure in a preset number of ƒ-stop increments. Even slight underexposure only increases noise in the image, while the “normal” meter-based exposure always can be made darker. Actual underexposure of the data provides no benefit. It would be equally useful if the camera manufacturers would build their light meters, LCD readouts and data handling based on RAW data. Until then, it's useful to understand the role of exposing for the RAW, linear data and, if possible, capturing as much data as a RAW converter can render. You may think you overexposed an image based on an initial rendering observed in a converter when, in fact, you might have up to a full stop or more data that can be normalized using appropriate exposure compensation.
In the final analysis, ETTR isn't about overexposure, but rather proper exposure, while avoiding true highlight clipping of linear-encoded data. This often isn't the exposure our light meters recommend.