High resolution satellite images are prone to geometric distortions. To correct these, the process of geometric correction becomes vital. Only knowledge of satellite altitude, attitude, position and the information of the digital elevation model (DEM) will not be adequate for the geometric correction requirements. Therefore we designed an algorithm for removal of geometric distortions in satellite imagery. In that a new method of geo-referencing called pixel projection method was applied along with selection of precise ground control points (GCPs). In pixel projection method vertices of remotely sensed image is geo-located based on ancillary data. For precision of GCP least square method was used to cater for instrument bias. GCPs were selected from Google Earth's software. Though with that approach precise geo-referencing of satellite imagery was achieved and a level-1 image was successfully converted to level-3 geometrically corrected image. In this paper we carried out residual analysis of our new proposed method. In first step an image to image matching was performed and their MSE (mean square error) was calculated. In second step 8 points in the original image and geo-referenced images were identified and their MSE was calculated. It is observed that with new approach of geo-referencing more precise geo-referencing has been done and image is found to be accurately geometrically corrected.
展开▼