See the accompanyng Instructional Video to this article

In a recent article I discussed how, using PANCROMA SPOT data sets could be combined to create false color and natural color composite images by combining the 20m resolution band files to produce a 20m resolution natural color composite image. It is also possible to go a step further, creating 10m natural color pan sharpened images from band files derived from the natural color composite image plus the panchromatic image. These take a little more effort than pan sharpening Landsat data. However the advantage of the exercise is 50% higher resolution in the resulting pan sharpened image. In fact, this technique will yield the highest resolution color composite result of any of the no-cost satellite image data sets. This article will walk you through the process.

I started with the S4_11730_4954_20060625_m20_X_utm11.tif data set downloaded from the GeoBase website. (GeoBase is the website that archives SPOT data for all of Canada). The first step is to create a natural color composite image using the three band files, in my case S4_11730_4954_20060625_m20_1_utm11.tif, S4_11730_4954_20060625_m20_2_utm11.tif, and S4_11730_4954_20060625_m20_3_utm11.tif. This can be done by selecting 'File | Open' and opening each of the band files in succession. Now select 'Band Combination' | 'Generate RGB Natural Color Composite from SPOT'. After a few moments a natural color composite image will be synthesized from the band files. You may need to use PANCROMA's image processing routines to improve the result. In particular, the 'SPOT Adjust' data group is particularly helpful in fixing one of the problems inherent with creating natural color images from SPOT band files. That is the fact that the green hues have too much red in them, often resulting in an unpleasing yellowish-green color to vegetation. The 'SPOT Adjust' image processing routine lets you add or subtract any combination of red, green and blue values from each pixel. Alternatively, you can remove red from just the green pixels by setting the 'Green Cutoff' and 'Red Cutoff Adjust sliders in this group. For example, if you select Green Cutoff=30 and Red Cutoff Adjust=30, then any pixel with a green value above 30 will have 30 units of red removed from it. The result will be that the green pixels will be adjusted selectively, without removing red tones from areas where you do not wish. My selected image was fairly well behaved so I accepted the unadjusted image for further processing. This image is shown at the upper right of the page.

The next step is to save the image. Then 'Close Graphics Windows and Reset' and reload the image that you just saved. Check the 'Generate Band Images' check box at the bottom of the main window. Now select 'Display One File' | 'Display One Color Composite Image' from the main menu. You should see three grayscale band files and the natural color RGB image displayed. Save the grayscale band files by selecting 'File' |'Save Grayscale Image' selecting either JPEG format. 'Close Graphics Windows and Reset' again. The RGB image can be discarded as it will not be needed any more for this exercise. It is shown to the right for reference.

Now load the panchromatic band, in my case S4_11730_4954_20060625_p10_1_utm11.tif. Display it by selecting 'Display One File' | 'Display One Grayscale Image' and then save it in JPEG format. 'Close Graphics Windows and Reset' again.

Now we are ready for the last step. If everything went well, you should have three JPEG grayscale band files derived from the natural color 20m RGB composite image and the original 10m panchromatic band, all in JPEG format. Now we will pan sharpen the images as normal by loading all four images (in the right order, band1 first and the panchromatic image last). Select 'Pan Sharpening' | 'HSI Transform'. Process the image.

My efforts is shown to the right. Clicking on the image shows the full resolution result. The considerable improvement in resolution compared to the base image is very apparent.

I chose a second data set for my next trial, also from GeoBase. This set, S4_11730_4954_20060625_m20_X_utm11.tif was more difficult to process than the previous set and allowed me to illustrate some features of the PANSAT image processing utilities. I started by opening the three band files and generating the natural color composite files as before. I then reloaded the color composite file and decomposed it to its band files, again exactly as described above. I also loaded the .tif panchromatic file and saved it as a JPEG as before.

I expected that I would have to make several adjustments using the Image Processing screen, so I created a set of subset images from the source .tif bands. Image processing iterations process much faster if you use subset images rather than full size images as the pan sharpened image must be recomputed each time you use a preprocessing utility. To do this I loaded the three band files and the panchromatic file, then selected 'Band Combination' | 'Subset Images'. After the 'Blue Band' image was generated I selected a representative subset area of the larger file and selected 'Enter'. I then saved the subset images as JPEGs by selecting 'File' | 'Save Subset Images' | 'jpg'. (Note that the images disappear when you select 'Enter'. They are redisplayed when you save them. This is an arrangement designed to conserve RAM for processing very large subset files.

Next, I generated an RGB composite image from subset bands 1, 2 and 3 by selecting 'Band Combination' | 'Generate RGB True Color Composite from ETM+'. Before doing so I checked the 'Maintain RGB Comparison Image' check box on the Main Window. Two images are then displayed. The image in the 'RGBComparisonForm' is persistent. It will remain displayed even after selecting 'Close Graphic Windows and Reset'. (It can be closed by selecting 'Close RGB Comparison Form and Reset' from the 'File' menu. You can use this image as a reference for adjusting the colors during the pan sharpening process.

The next step is to pan sharpen the subset images. After selecting 'Close Graphics Windows and Reset', load each of the subset band files and the subset panchromatic file. (They will be named with whatever base name you selected with the suffix 'bandx' where 'x' is either 1, 2, 3 or 4 for the assigned band number). Select 'Pan Sharpening' | 'HIS Transform' then process the file. After a few moments, you will see the pan sharpened image and the RGB Comparison image. These images are displayed to the right. You can see that the pan sharpened image has a more intensely green hue than the reference image. You can see exactly where the problem lies by dragging your cursor over both images. The R, G and B pixel values will be reported on the Main Window. The pan sharpened image has a lot more Green and somewhat more Blue than the reference image.

Now you can select 'Close Graphics Windows and Reset'. Reload the four subset images and pan sharpen again, this time checking the 'Activate Image Processing Routines' check box. When you process the image the 'Image Processing Data Input' box will be displayed. I removed some green and blue from the image by selecting 'Enable Channel Level Adjust' and moving the slider bars. I also selected 'Enable Pan File Adjust' and subtracted some brightness from the panchromatic image. I added some haze reduction by selecting 'Apply Haze Reduction' and moving the slider to 246. Finally, I checked the 'Save Settings' check box. This is very helpful when iterating image-processing adjustments, as all of your values will be saved between runs so that you can start exactly where you left off during the last iteration.

After a few runs, I was satisfied with my result. I then selected 'Close Graphics Windows and Reset' and loaded the full decomposed band files and the full panchromatic file (the one we saved earlier in JPEG format). I again checked the 'Activate Image Processing Routines' check box. When the image processing data entry form is presented, all of the settings have been preserved because the 'Save Settings' check box was checked. Now the same settings that were used to produce a satisfactory subset image can be directly applied to the full image without the need to write settings down. Simply click on the 'Process Image' button and the full sized pan sharpened image will be produced. It should look exactly like the processed subset image (except of course that it is larger).

Now you can save the file. If you feel that the image could benefit from a little post processing, select 'Post Process' | 'Color Adjust' from the menu. An image processing form will appear. The left hand image is the base image and the right hand image will preview your adjustments before applying the changes to the full image.

There are three sliders in the center of the form, one for each of the red, green and blue channels. They allow you to add or subtract up to 255 from each channel. Once you set the slider position, you can select 'Update Color Chip' to apply the changes to the adjust image on the right. You can modify the adjust image as many times as you like. Once you are satisfied with your settings, click the 'Apply to Image' button and the changes dictated by the slider positions and shown in the adjust image will be applied to the full image. Alternatively, select 'Cancel' to exit the form without making any changes to the image. If you make a mistake or don't like the result you must close or exit and start over.

Hopefully, this article has shown how to produce 10m resolution pan sharpened images from SPOT data. This is easy to do using GeoBase data. Hopefully Earth Explorer will someday include the panchromatic image with its image data so that this technique can be used for all of the United States as well.