Image fusion is a special case of information fusion, which more generally finds
applications in situation analysis and decision making. If data are collected as
signals quantized in pixels, an image comprises 2D pixels arranged as a picture.
As shown in Fig. 6.1, typical information fusion methods are dependent on where fusion is located in the processing chain. Image fusion techniques can be organized at different levels such as signal (from the sensor), feature (transformed), score (probability [not shown in the figure]), or decision (binary). Figure 6.1 illustrates how the processes are the same, i.e., data association, feature extraction, and decision declaration, but the difference is the order in which the processes take place and where they occur, i.e., at a level of individual sensor processing or at a central fusion location.
|