Panchromatic Image Sharpening of Landsat 7 ETM+

This tutorial provides steps to pan-sharpen the Landsat 7 Enhanced Thematic Mapper Plus (ETM+) multispectral bands with the 15-meter panchromatic band, without losing the RGB information from the original multispectral three-band 30-meter composite. The only requirement prior to following these steps is to ensure that all of the data are georeferenced or that all of the data cover the same geographic area. In this tutorial we will pan-sharpen multispectral bands using the panchromatic band from the same image.

Landsat 7 ETM+ data are acquired at three different resolutions. The multispectral bands (bands 1-5, 7) are collected at 30 meters, the thermal band (band 6) is collected at 60 meters, and the panchromatic band (band 8) is collected at 15 meters. The panchromatic band is a grayscale image that covers the red, green, and blue portions of the electromagnetic spectrum. The table displayed below provides more information about the characteristics of the Landsat 7 ETM+ sensor. Please note that while it is also possible to merge the multispectral data with the 60-meter thermal band, this procedure is not recommended because of the loss of data that would occur at the 30-meter scale.

Figure 1. Landsat 7 ETM+ Band Characteristics

The processes used to create pan-sharpened multispectral ETM+ imagery using both ERDAS IMAGINE and ENVI are shown below. (Please note that this does not signify an endorsement by the USGS EROS Center. Other remote sensing software packages may also contain similar functionality.)

USING ERDAS IMAGINE

There are a number of methods within ERDAS IMAGINE that enable pan-sharpening. For more information on the differences between the pan-sharpening methods, the preferred pan-sharpening method based on specific user needs, and the output differences resulting from the different methods, please refer to the ERDAS IMAGINE help documentation.

Two methods are provided below. The first method uses the Resolution Merge and the second method uses the Modified IHS (intensity, hue, saturation) Resolution Merge.

Resolution Merge method:

1. Open ERDAS IMAGINE
2. From the main menu, click Interpreter -> Spatial Enhancement... -> Resolution Merge...

Figure 2. ERDAS IMAGINE Main Menu

3. Within the Resolution Merge display box, click the folder button under High Resolution Input File: (*.img) and navigate to the location where the panchromatic band is saved
4. Once you have located the folder where the panchromatic band is saved, click the dropdown arrow next to Files of type and select TIFF
5. Select the panchromatic band (*_B80.TIF) and click OK
6. Next, click the folder button under Multispectral Input File: (*.img) and navigate to the location where the multispectral bands are saved
7. Again you will need to click the dropdown arrow next to Files of type and select TIFF
8. Select all of the multispectral bands (all bands minus *_B61.TIF, *_B62.TIF, and *_B80.TIF) and click OK
9. Next, click the folder button under Output File: (*.img) and navigate to the location where you would like to save the output data
10. Once you have located the folder where you would like to save the output data, click the dropdown arrow next to Files of type and select TIFF
11. Provide a file name for the output data and click OK
12. Next, under Method, select your preferred resolution merge option by clicking the radio button next to Principal Component, Multiplicative, or Brovey Transform (as stated earlier, please refer to the ERDAS IMAGINE help documentation for a description of these transformation options)

Figure 3. Resolution Merge Display Box

13. Under Resampling Techniques, click the radio button next to Nearest Neighbor, Bilinear Interpolation, or Cubic Convolution
14. For Landsat ETM+ data transformation you should not need to check the boxes next to Stretch to Unsigned 8 bit or Ignore Zero in Stats under the Output Options header
15. Under the Layer Selection header leave the Select Layers default value of 1:6 for Principal Component or Multiplicative methods; for Brovey Transform, select the three bands you would like to transform (please note that layer 6 will refer to band 7)
16. Under Data Type, click the dropdown arrow next to Output and select Unsigned 8 bit
17. Click OK to initiate the resolution merge (please note that it will take quite some time to run the resolution merge; you will receive an error message if there is not enough space to perform and save the output data to the specified location)

Figure 4. Job Status for Running Model

18. Once processing has completed, click Viewer from the main menu
19. Within the Viewer, click File -> Open -> Raster Layer...
20. Navigate to the location where the merged image was saved and change the Files of type to TIFF

Figure 5. Select Layer to Add Display Box

21. Click on the file to highlight it and click OK
22. To view the image, right click inside the Viewer and select Fit Image to Window
23. The pan-sharpened image will display

Figure 6. Selecting Fit Image to Window Option

24. To compare the pan-sharpened image to the original three-band multispectral 30-meter composite, click Viewer from the main menu again to display a second Viewer
25. In the second Viewer, click File -> Open -> Raster Layer...
26. Navigate to the location where the original multispectral bands are saved and change the Files of type to TIFF
27. Select all of the multispectral bands using the SHIFT and CONTROL key functions (do not click OK yet)

Figure 7. Selecting Multiple Bands

28. Click the Multiple tab at the top of the Select Layer To Add display box and select the Multiple Images in Virtual Stack radio button (it may take a few seconds, but the Proxy File option will appear)
29. To change the virtual stack default name, click the folder button next to Proxy File

Figure 8. Multiple Images in Virtual Stack Option

30. After renaming the file, click Save
31. Click OK
32. If given the option to build pyramids, proceed with doing so as this will speed up the upload time
33. To view the image, right click inside the Viewer and select Fit Image to Window
34. If desired, the RGB band combination can be altered by clicking Raster -> Band Combinations...

Figure 9. Selecting Band Combinations Option

35. Please note that layer 6 will correspond to band 7 (since band 6 is not multispectral)

Figure 10. Set Layer Combinations Display Box

36. After selecting the preferred band combination, click OK
37. To link the two displays, click View from the menu bar in one of the displays and select Link/Unlink Viewers -> Geographical
38. Drag the cursor over to the other display and left-click; the viewers are now linked (the images below show a comparison between pan-sharpened Brovey Transform data on the left and 30-meter data on the right)

Figure 11. Comparison Between Brovey Transform and Original Data

Modified IHS Resolution Merge method

*Prior to using the Modified IHS Resolution Merge, the ERDAS IMAGINE help documentation refers to Yusuf Siddiqui�s article �The Modified IHS Method for Fusing Satellite Imagery�.

1. Open ERDAS IMAGINE
2. From the main menu, click Interpreter -> Spatial Enhancement... -> Mod. IHS Resolution Merge...

Figure 12. ERDAS IMAGINE Main Menu

3. Within the Modified IHS Resolution Merge display box, click the folder button under High Resolution Input File: (*.img) and navigate to the location where the panchromatic band is saved
4. Once you have located the folder where the panchromatic band is saved, click the dropdown arrow next to Files of type and select TIFF
5. Select the panchromatic band (*_B80.TIF) and click OK
6. Next, click the folder button under Multispectral Input File: (*.img) and navigate to the location where the multispectral bands are saved
7. Again you will need to click the dropdown arrow next to Files of type and select TIFF
8. Select all of the multispectral bands (all bands minus *_B61.TIF, *_B62.TIF, and *_B80.TIF) and click OK
9. Next, under Resampling Technique, select your preferred resampling option by clicking the radio button next to Nearest Neighbor, Bilinear Interpolation, and Cubic Convolution
10. Under Hi-Res Spectral Settings, click the dropdown arrow next to Sensor and select Landsat 7 Panchromatic (leave the layer value and the Ratio Ceiling value as the defaults)

Figure 13. Modified IHS Resolution Merge Display Box

11. For Multispectral Spectral Settings, click the dropdown next to Sensor and select Landsat 7 Multispectral (leave the default layer values)
12. Next, click on the Output tab

Figure 14. Designating Output File and Location

13. Click the folder button under Output File: (*.img) and navigate to the location where you would like to save the output data
14. Once you have located the folder where you would like to save the output data, click the dropdown arrow next to Files of type and select TIFF
15. Provide a file name for the output data and click OK
16. Under the Data Type section, make sure that the Output is set to Unsigned 8 bit
17. For Landsat ETM+ data transformation you should not need to check any of the boxes listed under Processing Options
18. Click OK to initiate the modified IHS resolution merge (please note that it will take quite some time to run the resolution merge; you will receive an error message if there is not enough space to perform and save the output data to the specified location)
19. Follow Steps 18-37 above for viewing the pan-sharpened image and for comparing it to the original 30-meter data (the images below show a comparison between pan-sharpened Modified IHS data on the left and 30-meter data on the right)

Figure 15. Comparison Between Modified IHS and Original Data

USING ENVI

There are a number of methods within ENVI that enable pan-sharpening. This tutorial provides step-by-step guidance on how to perform the image transformation. For more information on the differences between the pan-sharpening methods, the preferred pan-sharpening method based on specific user needs, and the output differences resulting from the different methods, please refer to the ENVI help documentation. A header file is required to perform Gram-Schmidt Spectral Sharpening, PC Spectral Sharpening, or CN Spectral Sharpening. Because GeoTIFF files do not have a header file, and all Landsat data are currently distributed in GeoTIFF format, these methods will not be included in the following examples.

The following steps explain how to pan-sharpen Landsat ETM+ imagery in ENVI using either the HSV or Color Normalized (Brovey) methods.

1. Open ENVI
2. From the main menu, click File -> Open External File -> Landsat -> GeoTIFF

Figure 16. ENVI Main Menu

3. Navigate to the folder where the image files are saved, highlight all of the bands using the SHIFT key function and click Open; the Available Bands List display box will appear with the bands that you selected
4. Next, from the main menu, click Transform -> Image Sharpening and select either the HSV or Color Normalized (Brovey) method.

Figure 17. Image Sharpening Options

5. Within the Select Input RGB Input Bands display box, select the multispectral RGB combination that you would like to pan-sharpen and then click OK
6. Next, within the High Resolution Input File, select the panchromatic band (*_B80.TIF)
7. Leave the Spatial Subset as the default of Full Scene and click OK
8. From the next display box, select the resampling method by clicking on the dropdown arrow next to Resampling and select one of the following options: Nearest Neighbor, Bilinear, or Cubic Convolution
9. Next, click Choose and navigate to the location where you would like to save the output data
10. Provide a file name for the output data and click Open
11. Click OK to initiate the resolution merge (please note that it may take a few minutes to run the resolution merge; you will receive an error message if there is not enough space to perform and save the output data to the specified location)

Figure 18. Color Normalized Sharpening Parameters Display Box

12. After the resolution merge has been performed the new pan-sharpened RGB bands will be added to the Available Bands List
13. Ensure that the correct bands are displayed in the RGB fields and click Load RGB to display

Figure 19. Available Bands List

14. To compare the pan-sharpened image to the original three-band multispectral 30-meter composite, click on the Display #1 button within the Available Bands List and select New Display
15. A new set of image windows will appear and the display button will change to Display #2
16. Ensure that the radio button is set to RGB Color and select the three bands that were used during the pan-sharpening process and then click Load RGB
17. To geographically link the two displays together, click Tools -> Link -> Geographic Link... from the menu bar within one of the images
18. Within the Geographic Link display box, click the arrow buttons next to both displays so that they both change to On and then click OK

Figure 20. Geographic Link Display Box

19. The pan-sharpened image and the original 30-meter data are now linked (the images below show a comparison between pan-sharpened Color Normalized (Brovey) data on the left and 30-meter data on the right)

Figure 21. Comparison Between Color Normalized and Original Data


Download Adobe Reader

..