Epipolar Resampling of Pushbroom Satellite Images Using Piecewise Linear Implementation of Pseudo Epipolar Line

Abstract

This paper presents a practical and efficient epipolar resampling method for pushbroom satellite images. We proposed a new resampling scheme based on PEL (Pseudo Epipolar Line). The main concept of PEL is derived from the assumption that the projection line from the satellite image sensor is a straight line. In this context, the epipolar line can be assumed as PEL, which can be directly evaluated directly from RPC (Rational Polynomial Coefficient) model. The pushbroom satellite image sensor uses a linear array scanner for image capture. Since the resulting images are formed by collecting the one-dimensional line images with the sensor movement, the line assumption is best approximated in the vicinity of the area forming the one dimensional line image. In this regard, we devise a piecewise linear implementation approach of the PEL epipolar resampling method to best approximate the epipolarity model of the linear pushbroom image. In this new approach, whole satellite images are divided into small images (pieces), and epipolar resampling is performed on each individual piece using the PEL. This approach is very accurate because the epipolar lines in each small image are evaluated by conjugate points within the terrain corresponding to the small image. Experimental results show that highly accurate resampling was achieved when all the whole images are divided into multiple small pieces. Our study confirmed that the suggested method can be efficiently utilized in epipolar image generation.