Point spread function analysis for the altitude-dependency of multi-altitude digital image set.

摘要

The scale effect is a fundamental factor that needs to be considered in sensor design and data selection for different applications. Spatial scale effects in the application of remote sensing have been noticed and explored by scientists and various new methods have been applied to deal with their effects. However, most of the research only provide a way to deal with one or one set of problems instead of explaining their origin, especially for the image generating processing of remote sensors. In this paper, I develop an imaging model and explain how the scale effects are generated in the process of imaging by analyzing the model. Then, I test several hypotheses based on the model, which assumes a particular sensor response given by the reverse PSF (Point Spread Function), to explore the scale effects on spatial and spectral sampling (spatial resolution and recorded reflectance range). It is proved that the variations of the recorded signal and the power spectrum at each pixel are affected by illumination geometry, sensor PSF, ground cover types and their distribution, altitude, and bandpass. Therefore, the same ground scene can yield different images at different altitudes. Based on these conclusions, I analyze and explain the scale effects on image resampling algorithms and pixel unmixing algorithms and point out the inherent problem with them due to their neglecting the scale change in the imaging process. The conclusions from this research should be useful in designing new sensor and new image processing algorithm and selecting appropriate data for different applications.