There are a number of iPhone camera apps that promise to have superior performance in low light settings by using the iPhone’s accelerometer. The basic idea is very simple and intuitive. To quote from the Night Camera web site:
Due to low light condition, the shutter time is longer, and even small shaking from the tapping of the camera button will make the photo un-usable.
So, we created Night Camera, the app to prevent the blur at the first place. Using the built-in iPhone accelerometer, it automatically shoots the photo when it detects the iPhone being stable, so you have a real chance to get some good photos at night.
Sounds very compelling. Basically image stabilization (or rather image selection based on stability data) on inexpensive hardware.
There is one problem with this, which is that the iPhone according to this post uses a camera module that has a fixed 200ms rolling shutter. This seems very plausible, as most camera modules of cell phone use this technique. There also is plenty of photo evidence that this is the case. With a rolling shutter, the exposure time of the CMOS sensor is fixed. In other words, the claim that the shutter time is longer at night is wrong. Unless I am missing something, the makers of the above application at least don’t understand how the iPhone camera works.
Now it may be that in general the accelerometer could be useful for enhancing image quality by reducing movement. However anecdotal evidence shows no visible improvement, and the above app has only 2.5 start on iTunes.
I’d be shocked if the iPhone accelerometer has anywhere near the resolution you’d need to make IS work. Also, I would assume that the iPhone has no shutter at all but rather a sort of “soft shutter” that scans the sensor in a predefined rolling shutterish pattern.
If I recall correctly, an application like NightShot is not doing anything with the image. It’s NOT doing image stabilization. What it is doing is in software delaying the button press for the exposure to happen based on when the accelerometer reports that the phone is moving less. i.e. pressing the button doesn’t take the image, but it tells the app “please take an image, when the accelerometer reports the phone is mostly still” So the improvement comes from the delayed exposure (kind of like using a remote for a DSLR to overcome the vibrations from pressing the button) and not from any fancy image processing.
I believe you are correct that NightShot isn’t doing anything with the image, but simply wait for the right time to take a photo. My point is that at the very least the people developing NightShot don’t understand what they are doing. With a fixed length rolling shutter, the exposure time for a shot at bright daylight and at night is be identical. Let’s assume for a second NightShot improves picture quality, it should improve it exactly the same in bright sunlight (as the exposure is 200ms) and in complete darkness (as the exposure time is still 200ms). NightShot though specifically claims on their web site that their application works best at night. Unless I am missing something that doesn’t make any sense.
Now in theory it could be that they just improve any image, whether day or night. At least the anecdotal evidence of people that ran comparisons seems to indicate that there is no tangible difference though.
So long exposure shots would also be impossible with the 0.2 Second fixed exposure time?
I suppose the image would get too noisy as well due to the sensor being quite simple?