Additional information on the GLT and Super GLT methods

In ENVI, one method of georeferencing data is through the GLT and Super GLT methods. This Help Article goes into further detail on the processes that the GLT and Super GLT use.

What is a GLT file?
GLT stands for "Geographic Lookup Table." A GLT file contains map locations for every pixel of the image with which it is associated. The GLT file consists of 2 bands: one is the sample number, and one is the line number (both for the georeferenced version of the image).

What is an IGM file?
IGM stands for "Input Geometry." An IGM file contains the map information of the image, and this information is stored in two bands: one for X coordinates (e.g., longitude or northing) and one for Y coordinates (e.g., latitude or easting). Some datasets come with specific latitude and longitude bands, and it is these two bands that make up the IGM file. Additionally, ENVI generates IGM files for various sensors such as SeaWiFS and AVHRR. The input geometry (IGM) file itself is not geocorrected, but does contain the geolocation information for each original raw pixel.

What is a Normal GLT?
A normal GLT takes all the pixels in the input image (since there is a map x/y for each individual pixel) and places them exactly where they belong in the output image (rounded to nearest integer location in output pixel space). After this is done there are remaining "holes" in the output image where no input pixel landed. These holes are filled with the nearest valid pixel from the first step, which is essentially a nearest neighbor resampling.

The GLT file is a georeferenced image and is warped into the user-selected output projection and pixel size. The DNs in the GLT indicate the sample and line number of the pixel in the original input image that belongs at the given map location in the output image. If the GLT values are positive then an exact match from the input image was found (i.e., the input image actually contained a pixel for this map location). If the GLT values are negative then a nearest neighbor pixel is being used because a pixel was not actually contained in the input image for this map location. If the GLT value is zero then there was no nearest neighbor within 7 pixels of this map location in the input image. This is a very accurate method of georeferencing data because it places every pixel at its exact map location, avoiding the use of GCPs and polynomial warping.

What is a Super GLT?
A Super GLT is much slower than a Normal GLT, but its doing much more. The Super GLT process is doing an optimal radial resampling, which means that every input pixel within a given radius which contributes to the value of an output pixel is considered in a weighted fashion. In comparison, bilinear resampling arbitrarily averages a group of 2x2 neighbors; cubic convolution resampling arbitrarily averages a group of 4x4 neighbors. However, Super GLT averages every relevant pixel in the file within a given radius to come up with the best possible output pixel value.

The Super GLT method is particularly useful for aircraft data where a given point on the ground might be scanned multiple times by the sensor due to the roll, pitch or yaw of plane, or even an overlapping flight line, which means that an output pixel might be best represented by an average of three different input pixels which are located in three different locations in the input file. This is an example of why this method can take such a long time to process.

This process was developed to implement ENVI's particular georeferencing scheme and there are no references or papers associated with it.