Deconvolution is the process of removing signal degradation from recorded data. Signals have two primary ways of being damaged: either the signal is created or recorded incorrectly or the signal is interfered with as it travels from point to point. Any form of signal damage is called a convolution and deconvolution is the process of taking those convolutions out without damaging the original data. These processes are used heavily in signal processing, audio and video processing and seismology, but the underlying process is used in nearly all mainstream sciences.
Any disruption to a signal is a convolution. It doesn’t matter if the disruption is caused by another signal, a reflection of the original or even a faulty recording device. Small convolutions usually don’t disrupt the signal enough to worry about; these are often expected simply from the signal moving through space. On the other hand, large convolutions will make a signal unreadable and need to be removed.
The vast majority of deconvolution is determining what convolutions happened to the original signal. Once the exact convolution is determined, it is possible to modify the original to edit it out. The majority of the time, this simply means the signal is modified again by another convolution that is exactly opposite of the original disruptor. These two signals will cancel each other out and return the recorded information to its original form.
This process has a huge number of real world applications. Deconvolution is widely used as a method of correcting optical images to account for magnification distortion. When a lens magnifies an image, the image isn’t magnified evenly across the entire field. Even in high-end microscopes and telescopes, there is a very small amount of distortion. When an image is heavily magnified, either by looking at a very small or a very distant object, the distortion can radically affect the image. By applying an opposing convolution to the image, a much truer version is created.
This same technique is used in several other audio and visual fields to improve the signal strength and create more real recordings. In seismology, a signal is distorted by distance, medium and reflections of itself. All of these convolutions amount to a nearly useless signal. By using deconvolution to move backward through all of the distortions, scientists can learn more about what is happening at the point of the signal’s origin and what sorts of things exist between signal transmission and receiving.