Image Geometric Correction Methods

Introduction

Geometric correction is an important system that is normally performed on satellite images as a part of preprocessing a prior to the extraction of biophysical information. Geometric correction is utilized to remove geometric distortion of an image so that the pixels will be in a proper planimetric position. There are three types of geometric correction image-to-map, image-to-image and a hybrid approach that employees image to map and image to image. Geometric correction starts with a distorted image and a reference image. A collection of ground control points is collected onto the reference image to be interpolated onto the distorted image. Upon completion of the multipoint geometric correction the resulting image should be rectified

Goals

In this lab we were tasked to do two types of geometric correction image-to-map rectification and image-to-image rectification. We do both these steps with a reference image to create a planimetric image. The images that we had were skewed by internal error since the mapping satellite followed a NADIR path while the earth continued to move west to east creating a skewed version of the image. By using a reference image we can rectify this.

Methods

To create a rectified image ERDAS Imagine was utilized for this project. The first objective was to create a rectified image using a 1^st order polynomial equation in a map to image rectification. The distorted image is Chicago in digital raster format. The reference image is Chicago in USGS format. The distorted image and rectified image needs to be imported into ERDAS. Upon importation the multispectral raster tool needs to be selected. Control points need to be inserted into the distorted and reference image. The rectification can be a simple multipoint geometric correction this order of transformation is the first order. This means we need to enter a minimum of 3 GCPs.

                The GCPs need to be inserted into the distorted and reference image. Since we are using a first order transformation care needs to be taken to not insert more than 4 points to avoid crashing the tool. Upon entering the GCPs in both the reference and distorted image the Root Mean Square (RMS) error needs to be below 2.0 for correct placement, this sometimes require manually replacement of the GCP’s.

RMS error correcting

                The image can now be computed using the multipoint geometric correction tool. Since this image is using a 1^st order transformation we can use nearest neighbor interpolation. Once the image is interpolated we can see how the image is corrected into a geometric correct position.

geographically corrected on left, original on right

                The next step is to interpolate by image-to-image. This is done in the same way as the above image but with some differences. By using a reference image of Sierra Leone with a distorted image of the same area the image is rectified by using multispectral tool. For this image bilinear interpolation will be used, this means more GCPs are required due to the skewness of the image. After adding the 12 needed GCPs we are ready to interpolate. Upon running the bilinear interpolation we now have a correct image of Sierra Leone.

Bilinear corrected image on left with original on right

Results

The results of running geometric correction help to create spatially correct images in the form of interpolation. This data is useful to create correct images that are not distorted and can be evaluated without skewness.