A couple of articles ago I outlined how to access Geogratis Landsat 7 data and to combine bands 1, 2 and 5 ( or any three bands between 1 and 7) to produce a 30m true-color RGB composite. As discussed in that article, Lansat 7 files also include a 15m panchromatic band that consists of a 12,928 by 12,000 8-bit grayscale image. This image has the advantage of double the resolution of any of the other seven bands (and equal to the best ASTER bands) but also has the distinct disadvantage of being a grayscale image. I had seen references that a full (15m) resolution RGB image could be produced from the grayscale but at the time of the previous article I did not know how. Since that time I have figured out one way to do it easily and present it herein.
Beware: this method is not for the faint hearted. It requires simultaneously processing multiple image files that will together exceed 750MB in size. The procedure requires at least 256MB of RAM, and preferably 512MB or more. I managed to pull it off with my Emachines T4200 with 256MB but I had to shut down all other applications in order to complete the operation. Even so, processing required considerable memory paging and as a result was slow.
In this exercise, we will combine the information stored in three 30m Landsat bands (bands 1, 2 and 5 in this example) with the information stored in the single 15m panchromatic band in order to produce a 15m true color composite. There are several ways that this can be done. Each method will produce slightly different results. Typically, several are tried until a suitable image is produced. Most of the standard techniques require custom image processing software. However, one method can be done using only the world's-greatest low cost GIS application: (you guessed it) Paintshop Pro.
The procedure is as follows:
1) Load bands 1, 2 and 5 (or any of the other Landsat 7 bands except band 8) into Paintshop Pro.
2) Convert each image to a gray scale by selecting Colors | Grayscale.
3) Combine the three images into a single RGB composite by selecting Colors | Combine Channel | Combine from RGB. Place band 5 in the Red channel, band 2 in the Green channel and band 1 in the Blue channel.
4) Now decompose the resulting true color 30m RGB into three HSL images by selecting Colors | Split Channel | HSL.
5) This will produce three grayscale images called something like Hue*, Saturation* and Lightness*. Close and discard the image named Lightness*.
6) Now resize Hue* and Saturation* by 200% so that they are 12,928 by 12,000 pixel grayscales, i.e. so that they exactly match the size of the band 8 grayscale image.
7) Now open and load the band 8 15m panchromatic grayscale.
8) Merge the three grayscales by selecting Colors | Combine Channel | Combine from HSL. Put Hue* in the Hue channel, Saturation * in the Saturation channel and the band 8 grayscale in the Lightness channel.
9) The result will be a 15m RGB color composite about 443 million bytes large.
Thumbnails of my Hue4, and Saturation4 images produced from the three 30m bands and the L72003026_02620000827_b80 band 8 image are shown at the upper right. Directly beneath these is the 30m RGB composite that we produced in the previous article. Beneath this is a detail from the HSL composite produced in this exercise. The improvement in resolution will be immediately obvious if you click on the thumbnails and examine the full size images closely. Many airport details that were not resolved in the previous image such as runway markings, hangers, and buildings in the nearby town are readily visible.
This technique may not always produce an optimal image and you may wish to experiment with other techniques (that require some programming). However, I was quite pleased with how well this simple technique worked using our readily available tool, Paintshop Pro. If JASC knew what a powerful GIS application they had they would probably raise the price. Let's just keep this our little secret.