|
|
|
Brief description of the problem:
Many cameras have autofocus capabilities. They can fail, however, causing an unsharp, blurred image. A common situation is if we take a picture about two persons standing side by side each other, and the camera sets the focus to a tree between and far behind them. A similar situation is if we shoot video and somebody crosses the scene. The camera adjusts the focus to the nearest object, and the desired object becomes blurred. There are also situations where the autofocus fails to work (e.g. not enough light, too homogenous intensity of the object etc.). In this case the camera will move the lenses back and forth looking for a good adjustment (it is called hunting).
The image can be reconstructed if we know the distortion. Linear and space-invariant distortions can compensated. The optical transfer function should be identified, and the measurement has to be compensated for it. The distortion can be described by the so called point spread function, which is the intensity distribution of an infinitesimally small point source on the image plane. This distribution is actually the optical impulse response of the system.
A simulated example is shown here:
Original image (scanned picture) |
Distorted and noisy image |
Reconstructed image |