A recent project of mine has been to gather IR and VIS data from WeatherUnderground's online site for the purposes of creating fun and informational temporal GIFs. Now, I'd like to take it a step further and attempt to determine the cloud type at a particular location based on these two data sets (according, in part, to this paper). To do this, I will need a way to accurately compare pixel data across various regional data sets. For example, here are two images from the Central-South region:
Thankfully the state borders are clearly defined, so my first naive approach would be to manually create a coordinate system based on the edges of these states so that a cursor position would produce matching longitude/latitude coordinates, for both types of data sets. However, there are multiple regions I'd like to do this for (Central, Central-East/West/South) and realize this may not be the most efficient way to do so.
How can I efficiently and accurately compare the data from both of these image sets for the purposes of cloud classification?