Neural network architectures for median filtering
R.R. Koita, P.G. Poonacha · 2003
A median filter has been implemented by using a three-layer feedforward perceptron network. A feedback architecture, using the thresholding and stacking property of the median filter, has also been developed. An alternative perspective to the problem of determining the median is proposed. The new approach greatly simplifies the problem and results in an architecture in which the number of nodes in the hidden layers has a quadratic dependence on the number of inputs.>