Image fusion at the pixel level can be implemented as simply as an arithmetic average; however, there are many techniques that improve on the simple methods. For example, a weighted sum of two or more input images is a valid image fusion. Principal component analysis (PCA) can be applied to decide the weights of input images (i.e., which input is more significant). In these methods, input images are directly combined in the spatial (pixel) domain. This chapter fully describes two multiscale (also called multiresolution) fusion techniques: pyramids and wavelets. The multiscale fusion processes are usually performed in the transformed domain. Color image fusion and multi-image (three or more) fusion are introduced as examples. The chapter also highlights recent techniques that extend the wavelet concept, including bandelets and contourlets for image fusion. Finally, a set of fusion examples is presented for multimodal and multiscale image fusion. |
|