Neural Network Correction Algorithm Based on Neighborhood-Weighted of Infrared Image

摘要

To restore original infrared image with non-uniformity noises, adaptive non-uniformity correction algorithm has become a hot research currently. Neural network algorithm based on human visual imaging principle has a unique advantage in this regard, and this paper studies neural network correction algorithm of infrared image. Based on the analyzing of current neural network correction algorithms, the paper puts forward a new neighborhood-weighted neural network algorithm, which has a good ability to eliminate horizontal and vertical banded-pattern noises of original infrared image with non-uniformity noises. Also, experiments show that the improved algorithm has good effects for image restoration.