Automated assessment of NIIRS and GRD of high resolution satellite images through edge profile analysis of natural targets

초록

As high resolution satellite images become more and more popular, image quality has been one of the key concerns for scientific and industrial applications. At previous JACIE workshops, several authors discussed the automated estimation of spatial resolutions of high resolution satellite images. Notably, Pagnutti and Ryan at JACIE 2010[1] discussed the possibility of automated estimation of RER (Relative Edge Response) and MTF (Modulation Transfer Function) ratio at a cutoff frequency from edge profile analysis of natural targets. In this paper, we propose to use NIIRS (National Imagery Interpretability Rating Scale) and GRD (Ground Resolvable Distance) as measures to describe spatial resolution of high resolution satellite images. They can be estimated without Fourier Transformation of edge profiles and hence are free from accuracy problems related to interpolation and curve fitting. We also propose to use natural targets, instead of artificial ones, for edge profile analysis and NIIRS/GRD estimation. We introduced this concept firstly at IGARSS 2008[2] and discussed further at ASPRS 2010[3]. We have achieved automated estimation of NIIRS and GRD from natural targets by developing a reliable edge profile analysis technique and by designing systematic and efficient edge selection scheme. At the workshop, we will show that automated edge analysis from natural targets have shown comparable performance to manual edge analysis from natural and artificial targets. We will also compare the NIIRS and GRD estimated by the proposed technique with the true values.