A Multiwindow Approach for Radargrammetric Improvements

Abstract

Abstract-The most intuitive way to extract depth information from remote sensing images is stereogrammetry, in which a digital elevation model (DEM) is achieved by computing stereoscopic radar images. When only the amplitude of the radar images is considered, this computation is called radargrammetry. The main idea of which is to match stereopair radar images in order to create a disparity map from one image to the other and, finally, to compute the elevation. Therein, we present our studies on the extraction of 3-D information from radar images. We examine a way to produce a DEM of a challenging area of the French Alps. The central issue of this paper concerns improvements for radargrammetric synthetic aperture radar image processing for high-relief reconstruction, and we focus on the matching step, which is one of the most important points of the radargrammetric processing. Thus, we propose original methods using different correlation windows. On the one hand, we take the advantages of a multiwindow approach to combine relevant information by multiplying the correlation surfaces obtained for each correlation window size during the matching operation. On the other hand, the second improvement is based on the expansion of windows on foreshortened areas, particularly because of the side-looking radar view. These methods allow us to achieve reliable image matching and to improve the accuracy of the DEM.

A Radargrammetric Approach for Spaceborne Transmitter-Stationary Receiver Bistatic Sar

“…more similar images, in the later configuration the intersection angle between the two image pairs is much larger, which increases the accuracy of radargrammetric products. Depending on the application, one configuration could be preferred over the other [8].…”

“…The main disadvantage of this type of algorithms is that the considered images should be quite similar in order to make a correct match. This is why they are usually used with figures in epipolar geometry, or when the variation of the incidence angles is small [8,10].…”

“…In our study, we have implemented a classical window matching technique, as described by [8], based on maximizing the zero normalized cross-correlation coefficient (ZNCC). The use of this algorithm is legitimate since the difference between the incidence angles of the two orbits is quite small (roundabout 9 degrees).…”

Similarity measures of full polarimetric SAR images fusion for improved SAR image matching

“…It is defined by the following equation: (7) where = matching window = window size One of the most commonly used SAR image matching similarity measures is Normalized Cross-Correlation (Méric et al, 2011) which is based on an area between two or more SAR images. And it can be stated as follows: (8) Where R = reference image M = matching image = parallax of range direction = parallax of azimuth direction The computation of NCC values in each polarimetrc band such as HH, HV, VH and VV can be performed similarly.…”

The Epipolarity Constraint in Stereo-Radargrammetric DEM Generation

“…We furthermore incorporated dedicated improvements, like the combination of different normalized cross-correlation kernel sizes as proposed by Meric et al [23] and the prediction of image disparities using the physical SAR sensor model and a coarse DEM, as proposed by Raggam et al [1].…”