Analyzing Multi-sensor Data Fusion Techniques: a Multi-temporal Change Detection Approach

An inevitable tradeoff between spectral and spatial resolution exists in remote sensing systems and it is essential to integrate information from systems with complementary characteristics to improve data interpretability. Panchromatic (PAN) sharpening is a data fusion technique that merges a high spatial resolution PAN image with lower spatial resolution multispectral images. The resulting merged data, which inherits the spectral characteristics of the multispectral images, has the spatial resolution of the PAN image. This paper presents a comprehensive assessment of PAN sharpening techniques. Numerous studies have investigated the use of sharpening techniques for improved machine interpretability and evaluated the sharpening performance using univariate statistical measures. However, studies have not explored the interaction between sharpening and change detection. Since the ability to repeatedly acquire images over a specific ground area for detecting changes is sensor platform dependent, the significance of multi-sensor image change detection has increased in recent years. In fact, change detection can serve to be an effective measure for evaluating the performance of sharpening methods. By employing an image acquired in a spectral range most pertinent for detecting changes in a specific material; change detection can highlight the inefficiency of methods in sharpening individual images. Hence, the effect of sharpening on change detection is exploited in this paper and the experimental results underline the significance of change detection as an effective measure for evaluating the PAN sharpening methods.