We illustrate here two applications, edge detection and impulsive noise removal.
In the first case the CNN was designed from a truth table, whereas in the second it was trained from an example.
Edge Detection
Flowers
Cameraman
IC
Coins
Impulsive Noise Removal
Pedro
Blood
Asado Argentino
Flowers
Bacteria
Blood
Impulsive Noise Removal Quantization
If you quantize the S-CNN coefficients with only two levels (0,255) you obtain the following images
Quantized version of the Gene (blue = 255, white = 0)
Layla
Pedro
Asado Argentino Bacteria
Flowers