Nonlinear Filter

What is a Nonlinear Filter?

A nonlinear (or non-linear) filter is a filter whose output is not a linear function of its input. In short, the input values produce output values that are not linear translations. It may be as simple as an exponential increase, as in a squared, or cubed output. Nonlinear filters are useful in image processing and convolutional neural nets.

How does a Nonlinear Filter work?

While the name can seem daunting, the principle idea behind a non-linear filter is fairly simple. Take, for example, an electrical device. At any point, the output voltage is a square of the input voltage. Alternatively, audio signals often have to be translated when transmitted between points. All radio transistors convert audio signals from kilohertz to gigahertz, and the digital equivalent converts analog audio signals to binary values. Nonlinear filters are incredibly useful for applications involving data scalability and shift invariance.