Landsat image pre-processing in ArcGIS - tools for seamless mosaicking

Mosaicking adjacent Landsat tiles often produces visible seam lines at the boundary between the two scenes. The Landsat Toolbox for ArcGIS deals with this problem by selecting non-cloud, non-shadow, and non-snow pixels in the overlapping portions of the scenes and then performing linear regression on each band. To do this the user must first select a "master" or reference image and a "second" image. The values in the second image are adjusted to match the first image. Although this method is relatively simple it is not part of the standard mosaic tool in ArcGIS. As evidenced in the figure below this method is quite effective at removing seam lines producing a visually coherent mosaic.

The upper image contains portions of two overlapping scenes mosaicked using the standard mosaic tool. The lower image was mosaicked using the Landsat Toolbox for ArcGIS.

Landsat image pre-processing in ArcGIS - tools for topographic correction

Correcting Landsat imagery for topographic effects is challenging at best and ignoring the differential illumination resulting from topography can lead to some pretty misleading results, especially when it comes to change analysis. Topographic efffects arise not just from shadowing, but also (more importantly) from different angles of the ground relative to the sun angle. When the sun is directly overhead pixels will be much brighter than when the sun angle is at a much lower angle. In our lab we've used a variant of the empirical-line method for removing topographic effects on Landsat image. In the Landsat Toolbox for ArcGIS 10.1. First, a Landsat metadata on sun angle and azimuth is used to generate a hillshade (illumination) raster that mimics illumination at the time of satellite overpass. Then each band is extracted and a linear regression is used to predict the reflectance as if each pixel were illuminated the same. We've found this method to be simple and quite successful at reducing the differential illumination effects in an image.

The images below illustrate some mountainous terrain before (left) and after (right) topographic correction. You can see that the terrain appears flat aster topographic correction.

New tool - Landsat image pre-processing in ArcGIS - Part I

Landsat is powerful resource for measuring changes on the Earth's surface over the past > 30 years. However, ArcGIS users lack image pre-processing tools available in remote sensing packages, such as ENVI and ERDAS. Through the years I've trialed and error-ed with different image pre-processing workflows with varying success. Some workflows actually ended up making the data worse off than when I began! Recently I decided to post the Landsat Toolbox that I've developed to facilitate my own image processing - http://www.arcgis.com/home/item.html?id=a60b0120a79f45ae990bb85f4d12edee .

The Landsat Toolbox for ArcGIS provides many basic preprocessing tools that can be used to help facilitate change detection and vegetation dynamics studies. This toolbox lessens the need for commercial remote sensing packages, such as ENVI or ERDAS, and brings some image processing functionality directly into ArcMap. Image pre-processing involves steps that may be under-appreciated by some GIS analysts, but are nonetheless important for ensuring reliable outcomes. This toolbox contains tools to do the following:

1) Convert raw DN values to top-of-atmosphere reflectance
2) Perform radiometric normalization using user-selected pseudo-invariant pixels
3) Perform topographic corrections using a digital elevation model
4) Mosaic adjacent scenes using linear regression to ensure a smooth edge-match

Many of the tools in this toolbox require fmask or fmask for R to perform cloud, cloud shadow, and snow masking prior to running. However, you could also do the masking manually by setting any value that you wish to remove to > 0.

https://code.google.com/p/fmask/
https://bitbucket.org/droletg/fmask