Fusing images of same view from different sensors for night sight

摘要

The process combines images (X1, X2) of the same view, which are input from sensors which work in different regions of the electromagnetic spectrum, in order to produce a fused, geometrically-aligned image (Y). A classifying device which can be trained, is used for the joint evaluation of the features of those pixels from the different input images, which correspond locally. The device selects an optimal fusion operation for each pixel, and may be a neural network which can be trained. The training may be derived from image fusion masks selected by human experts.