ABSTRACT
Data fusion techniques can be classified into two different categories. One is the spatiotemporal domain fusion, and the other is the transform domain fusion. The former may include Intensity-Hue-Saturation transformation, principal component analysis, and wavelet transforms; the latter can be further classified into different categories, including neural networks-based fusion algorithms, probabilistic fusion algorithms, statistical fusion algorithms, and unmixing-based fusion algorithms. Some hybrid approaches exist to integrate the advantages from different data fusion methods. In this chapter, to empower the system design of these algorithms and optimize the efficiency and efficacy of Earth observation and environmental monitoring, three benchmarking spatial domain data fusion algorithms will be further delineated in terms of their mathematical constructs. The specific algorithms that will be discussed are the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) (i.e., a statistical fusion algorithm), Bayesian Maximum Entropy (BME) (i.e., a probabilistic fusion algorithm), and Deep Neural Network (DNN)-based fusion framework (a neural networks-based fusion algorithm). STARFM, BME, and DNN have been widely applied to different real world applications.
