Image fusion for a digital camera application

Image fusion techniques are investigated for blindly constructing an improved image from several distorted visual images when the distortion is mainly due to parts of the image being out of focus. A generic framework for image fusion based on multiscale decomposition is proposed. Different image fusion approaches described by this framework are carefully evaluated and compared. Misregistration problems are also discussed. Experiments show that a region-based image fusion scheme using a shift-invariant wavelet transform provides the best performance, for the digital camera application, among the different image fusion approaches investigated.