Improved Methods for Matching Process between SAR Image Pairs for Both Long and Short Baselines
Eftekhari, Akram1; Ghannadi, Amin1; Motagh, Mahdi2; saadatseresht, Mohammad1
1University Of Tehran, IRAN, ISLAMIC REPUBLIC OF; 2GFZ, GERMANY

SAR image matching is one of the most important processing steps for Digital Elevation Model (DEM) extraction using interferometry and radargrammetry methods. Because of radiometric distortion as speckle noise, layover, foreshortening and shadow, the matching processing is more complicated in SAR images than optical images. Limiting search area is one the methods to improve SAR images matching in the Area Based Matching (ABM) method. In this article we use epipolar line geometry for limiting search area of Terra-SAR-X image pairs during the matching. Epipolar line in the slave image is determined by Rang and Doppler (RD) equation and by orbital information, but solving direct and indirect RD equations for all pixels in master images is very time consuming processing. Therefore we employ some mathematical functions including linear, bilinear, and quadric functions for determining relationship between master image coordinate and epipolar line parameters in the slave image. The results show linear and bilinear equations can relate image coordinate in master and epipolar line parameter in slave image by accuracy better than 0.3 pixels. Thus with only 4 points in master images that were matched with corresponding points in slave image, we can find epipolar line in the slave for all pixel in master image.