Push processing (pushing, uprating) is a well known film photography era technique, in which the photographer intentionally underexposes the film then compensates for the underexposed film by over-developing it in the processing lab. Typically this is done by telling the camera the loaded film is rated for higher speed than it actually is. The purpose of using this technique is to obtain the needed shutter speed to avoid blur caused by camera shake or freeze motion.

With today’s digital technology, the sensitivity of the sensor (ISO) can be easily changed at any time to obtain the desired shutter speed. In the film days, it would have meant exchanging the film in the camera with another roll that is rated at a higher speed. Is there any more incentives for underexposing then compensate it in post-processing (the equivalent of a film processing lab) with today’s digital photography? After all, many people suggest that the shadow needs to be properly exposed to avoid noise. Before answering this question, let’s look at the results of my recent experiment.

I took the following photo at ISO400, with a camera suggested correct exposure of 1/100s @ f/8. I then set the ISO of the camera to 100 and took another photo while keeping the shutter speed and aperture the same. The result is shown in the 2nd photo below. The foothill shows nice saturated color but the shadow areas show little details.

ISO400 1/100s @ f/8

ISO100 1/100s @ f/8

During the shooting process, I have the camera (Nikon D200) mounted firmly on a tripod. Set the shooting dial to Mup (mirror lock-up) and used a remote cord to release the shutter. The focus is set to manual to avoid any possible variations. I also turned off the high ISO NR (Noise Reduction) since the high ISO NR kicks in at ISO as low as 400 in the D200. All photos were shoot in 12-bit compressed RAW format.
The post-processing was done in Nikon Capture NX v1.2. For the first photo, no additional processing was done. It was just simply saved as jpeg at the 100% quality. For the 2nd photo, I saved one copy as jpeg just like the first one. I set the exposure compensation slider in Base Adjustment/Raw Adjustments to +2EV, then saved the third image as jpeg at 100% quality as shown in the photo below.

ISO100, 1/100s @ f/8, pushed two stops

This pushed photo is very similar to the first one and I found it very difficult to see the difference by going back and forth with them. So I did a pixel by pixel comparison using ImageMagick. The image below represents the pixel by pixel difference. The brightness of the pixels represent the difference: darker pixel represents smaller difference and brighter pixel represents larger difference.

Difference of the 1st and third photo

To make the difference easier to see, level adjustment was made to normalize the histogram. Te result is shown below. The difference in sky is very minimum (the color streaks are interesting). There are some clearly visible differences in the shadow areas of the trees, the shadow of the ridges over the hill, and in the water.

ISO100 1/100s @ f/8

To find out what the differences really are, I cropped five 450×450 squares out of the image as illustrated in the next image. To view the differences, simply scroll down and move your mouse in and out of the images labeled as “Site #1″, “Site #2″, etc. The images shown with the mouse placed over it is from the pushed photo.

Crop locations for rollover comparisons below.

Site #1.

Site #2.

Site #3.

Site #4.

Site #5.

Most of the differences are not related to image quality. At site #1, the shadow of the ridges moved slightly as the Sun is setting. At site #2, the tree branches moved due to wind or birds. At site #3, something disturbed the water. At site #4, the shadow moved up slightly as the Sun was setting. At site #5, there are some subtle differences in details and colors. If you look closely, the pushed image appears to have slightly more noise but also contains more details.
The answer to the question I asked earlier is “yes”. This test demonstrated that you can gain two stops of shutter speed by intentionally underexpose 2 stops (used the shutter speed and aperture metered for ISO400 at ISO100). It is really difficult to find any substantial differences. This example may not have been the perfect one. It would have been better if there were some clouds in the sky. It is reasonable to believe the highlights would be preserved better by underexposing. This example has shown that the dark shadow areas contain amazing amount of details you can extract during post processing. Experience might tell you that blown highlights are much more difficult or impossible to recover.