Photophys.com: The Science of Photography

I am really concerned about the misuse of terms, and I will rant about that from time to time. Now it’s High Dynamic Range (HDR). When a photograph is displayed with unnatural gray sky, over-saturated colors, or pumped-up shadows, someone will always say that it has the HDR look. Give me a break! HDR is getting a bum rap. Scenes in nature often have high dynamic range. For example a bright day with gleaming clouds and dark shadows may exhibit 14 stops (EV) of luminance contrast. Now for the first time HDR techniques permit photographers to capture the full dynamic range and color gamut of a scene. This gives the knowledgeable photographer the opportunity to produce a low dynamic (LDR) image on a computer screen or paper that is as limited as a single photograph, an improvement on the single shot, or an art work with extreme saturation contrasts. That is to say, HDR really means “better than photographic.” Of course, I am not the only one defending the good name of HDR, see: http://www.luminous-landscape.com/essays/hdr-plea.shtml

So what has gone wrong with so called “HDR” images. I think that either the photographer failed to make a full HDR capture or the tone mapping step was botched. Let’s look at these steps separately.

HDR Capture: With current image sensors, full HDR capture requires that a set of images be made. This set must include images that properly expose the brightest parts of the scene, the darkest parts of the scene, and everything in between in steps of two F-stops (RAW) or one F-stop (JPG). Some experts advise that each pixel should be fully visible (not clipped) in at least two images. This means that one should check the exposure of the brightest and darkest parts with spot metering before planning a sequence of shots. One may be able to get away with three shots; but often five, seven, or nine shots will be required. Photographers who blindly make three shot brackets every time regardless of the conditions may sometimes improve on their standard images but often are not really getting HDR capture. Christian Bloch discussed HDR methods in detail in his wonderful book: The HDRI Handbook.

Tone Mapping: With photographic methods and HDR in particular there is the problem of squeezing the captured dynamic range into the limited dynamic range of computer screens (10 stops?) or paper (7 stops). With film the compression was handled automatically by the dyes in slides and print paper. Now the HDR photographer has great freedom in mapping the tones onto display devices. Programs such as Photomatix combine images into an HDR file and then tone map to provide a visible preview that can be manipulated to taste. It is very convenient, but also very easy to “get it wrong.” The white and black points need to be carefully set, and then the strength, saturation, details enhancer, etc., adjusted to taste. Unpleasant results can easily be obtained, especially if the initial image set is not complete. There are many HDR programs available now and a list can be viewed at www.hdrlabs.com. One program I find particularly powerful is HDR Expose (formerly HDR Photostudio). This program permits one to carry out all adjustment while remaining in 32 bit mode and to save the HDR file in an efficient format. HDR Expose handles tome mapping with the complete set of visible colors unlimited by RGB color spaces so that colors can be maintained.

There are two other points to get out of the way. Why not just process a single exposure in Photoshop or Lightroom so as to darken the highlights and brighten the shadows? This approach will work if the dynamic range of the scene is limited and the image is properly exposed. However it cannot handle blown out high lights and black shadows. Dark shadows can be brightened, but noise will become evident. What about processing a RAW image twice, once for the highlights and once for the shadows and then blend them? In principle this method cannot increase the dynamic range; however it may improve the handling of noise. Some photographers use the double processing method and claim an improvement. Full HDR is, of course, much better; but DR capture is impossible in the presence of rapid motion.

Finally, why do we have to do all the HDR stuff to get an HDR image. Why not have sensors or special in-camera capture sequences that will give HDR without pain (or thought). I expect that automatic HDR is in our future. One obvious method would be to offer a very high pixel density sensor in which some pixels are covered with neutral density (ND) filters. Imagine a sensor in which each of our current pixels is replaced with four pixels. One would be unchanged, and the other three would be covered with two, four, and eight stop ND filters in sequence. Each image then would contain all the necessary information for the generation of a 32 bit HDR image. So the future will give us HDR with full color gamut and many other features transparently. The days of LDR capture will fade in our memory just as photographic film has.