Noise Removal Based on Artificial Intelligence Synergy Methods for Surveillance Scenarios

Surveillance footage pose particular difficulties in terms of low-quality either in resolution due to high compression or sensor inconsistency. Latter ones could generate several types of noises such as salt and pepper noises. In this article we apply and evaluate an innovative method for salt and pepper noise removal on surveillance images. The method combines synergy of neural networks, evolutive and statistical methods. The Pulse Coupled Neural Network (PCNN) is used to detect the pixels affected by noise whose parameters have been optimized using a Genetic Algorithm. The pixels that are affected by noise are corrected using a Gaussian Kernel. The proposed method proves to be high efficient o n detecting a nd removing s alt a nd pepper noise from images without affecting surrounding clean pixels. The method has been tested on images from real-world CCTV cameras. The examples from Scotland Yard database show how the filtered images with o ur proposed method retain fine details of the scene even if the noise level was high.