Multi-frame image fusion using the expectation-maximization algorithm

A multi-frame image fusion scheme is proposed to fuse visual and thermal images for night vision applications. While many previous image fusion approaches perform the fusion on a frame-by-frame basis, this method considers optimum use of neighboring frames to incorporate temporal as well as sensor fusion. This fusion scheme is based on a statistical image formation model. The multiple sensor image frames are described as the true scene corrupted by additive non-Gaussian distortion. The expectation-maximization (EM) algorithm is used to estimate the parameters in the model and to produce the final fused result. The experimental results showed that the EM-based multi-frame image fusion scheme has significant advantage in terms of sensor noise reduction.