ABSTRACT
Image fusion methods based on multiscale transforms are a popular choice. Wavelet theory has emerged as a well developed rapidly expanding mathematical foundation for a class of multiscale representations. Based on the persistence property of the wavelet coefficients, we organize the wavelet coefficients of a source image as a forest of quadtrees. Each coefficient represents a node in one of the quadtrees. The trees are rooted at the wavelet coefficients in the high-frequency bands in the coarsest scale. Some might suggest that in night vision applications, the most information is contained in the thermal images, with complementary information from the visual images. The dependencies can be described using the statistical properties called clustering and persistence. Clustering is a property that states that if a particular wavelet coefficient is large then adjacent coefficients are very likely also to be large.
