Pre-processing of remotely sensed images is generally composed to two distinct processes:

1. Image enhancement or radiometric correction
2. Georeferencing or geometric correction

You have already performed a simple image enhancement exercise in Lab 5, when you 'destriped' the SPOT image (Exercise 4-2 Image Restoration and Transformation). Such enhancements are often necessary since, over time, many sensors develop 'glitches' that, while not rendering the sensor incapable of producing any data, none-the-less result in images that have variations in the data that are unrelated to the target. That is, there may be variations in the radiometric values that are unrelated to what the sensor is seeing (i.e., they are unrelated to the target), but instead are a result of a mis-function in the sensor itself. Unless we remove these 'internal' variations, any classification we perform on the image will conflate these (meaningless) internally-derived variations in the digital numbers with the (meaningful) externally-driven variations in the spectral response patterns that we are interested in.

In addition, when combining multiple images together (a process called mosaicing) we often need to adjust the overall characteristics of the digital numbers in one image to match those in another (e.g., because of the sun angle, one image may be 'brighter' overall than the other image--unless we adjust the DN values to match the two images will appear to be different, even though they physically are the same).

Georeferencing is a necessary process that all remotely sensed images must undergo, if the images are to be used in conjunction with other data sources. Only by transforming the raw images into a common coordinate system are we able to combine the imagery from one sensor with images collected by another sensor, and to combine remotely sensed images with all of the other spatial datasets being produced (such as digital elevation models, transportation maps, etc.).

Some useful links:

• A USGS page that has an interactive destripe applet (it also provides some useful overviews of other spatial filtering techniques).
• Another page that documents the banding problems often observed in Landsat images.
• A page that describes the pre-processing that all Landsat images undergo.
• High dynamic range photography (aka visual sensitivity versus remote sensing).
• A brief discussion on atmospheric and topographic corrections for satellite imagery.

Text: Chapter 3: Spatial Referencing; Chapter 5.3: Radiometric Corrections; Chapter 6.3: Two-dimensional Approaches; Chapter 8: Digital Image Interpretation

I have photocopied the Appendix on Rectification and Georeferencing of Optical Imagery from the text Remote Sensing for GIS Managers (Stan Aronoff, 2005; ESRI Press; now out of print) that covers this topic in some detail.

ESRI has a useful help file on georeferencing a raster dataset. A blog that includes a discussion on image destriping (part of the blog on Geometric and radiometric corrections).