Hebbian Neural Network Based Algorithm for Video Processing
Ismail A. Ismail,
Nabawia A. El-Ramly,
Hani M. Ibrahim
Traditional video processing methodology consists in transforming an original video in a processed video, using a well-known transformation function. This function is coded into the application. Our goal is to make a system able to perform image processing tasks without knowing the transformation function as traditional methods. We have used Artificial Neural Networks-ANN (Hebbian learning rule) in order to achieve this goal that is easy and flexible. The main advantage of this approach is that we havenít used predefined algorithmic functions, but video visual features learning. However, in the proposed method each frame is processed independently from the others. This studies goal is to show the viability and usability of Neural Network training, in creating "adaptive intelligent filters", using Hebbian learning rule. The comparison between the proposed technique and the temporally average filter is included.