My previous articles describing Landsat 8 image processing techniques outlined methods for using PANCROMA™ utilities for transforming raw Landsat 8 band data into visually pleasing RGB color composite images. The next step is to pan sharpen the image processed RGB composite image to improve its resolution.
The processed Landsat 8 bands 2, 3, 4 RGB color composite image that I demonstrated in my last article is shown below. This image will be the starting point for this exercise.
Landsat 8 bands 2, 3, 4 RGB color composite, processed.
As we saw in the previous article, producing visually pleasing RGB images takes a bit of work. The most important image processing elements that are needed to improve Landsat 8 imagery are transforming band files in Digital Number (DN) format to top of atmosphere (TOA) reflectances and then performing a proportional image stretch. Once you have produced an RGB color composite image that satisfies you, pan sharpening the image is actually pretty easy.
The first step is to decompose the RGB color composite image back to its corresponding grayscale band images. Then the band images can be pan sharpened using the unprocessed Landsat band 8 image. In order to do this, open the processed RGB low-resolution color composite image. You may have saved this image in PNG or TIFF format after previous processing steps. This is done by selecting 'File' | 'Open' and then selecting the RGB color composite file.
Now, check the 'Generate Band Images' check box at the lower left of the Main Window.
PANCROMA™ main window showing grayscale check box.
Then select 'Display Files' | 'Display Single Color Composite Image'. Your RGB color image, the three corresponding blue, green and red band images and a thumbnail image will be displayed. (Note: if you do not wish to display the color image, un-check the 'Generate RGB Color Composite Image' check box. It is not necessary to display the color image for this exercise.
Grayscale and RGB images.
Now select 'File' | Save Grayscale Image' | 'GeoTiff'. You will be prompted for a base file name. Input your file name. PANCROMA™ will automatically append suffixes to the base name and save the three grayscale images as '[baseName]band1, [baseName]band2 and [baseName]band3. Now select 'Close Graphics Window and Reset'.
The next step is pan sharpening. Open the three band files that you just saved plus the unprocessed Landsat 8 panchromatic image by selecting 'File' | 'Open' and selecting the file names. Select 'Pan Sharpening' | 'HSI Transform' | 'Standard Four File'. Accept all of the defaults when the subsequent data entry screens appear. After a few moments, your pan sharpened image will appear. The result of my pan sharpening is shown below. (Note that it was not necessary to perform an pre processing of the panchromatic image in order to perfectly reproduce the RGB color tones in the composite image for this example).
Pan sharpened image.
I decided that the image would benefit from a bit more stretching. To do this I first saved my pan sharpened image in PNG format by selecting 'File' | 'Save RGB Image' | 'PNG'. After saving the file I selected 'Close Graphics Image and Reset' I then re-opened the PNG file and selected 'Pre Process' | 'Histogram Stretch' | 'Proportional Stretch' | 'Color Image'. The Histogram Form became visible. I entered a scale factor of 2.4 for my image and clicked 'OK'. (Note: It may take a couple of attempts to get the scale factor exactly right). The result of stretching is shown below. This image has virtually the same color tones as the RGB composite image, but of course has doubled the resolution from 30m to 15m.
Pan sharpened stretched image.
This article has shown how to transform raw Landsat 8 band data into visually pleasing pan sharpened color images. The image processing requires a bit of technique, but the quality of the Landsat 8 imagery has lived up to expectations and it is possible to get very good results with a bit of effort.