Landsat and SPOT^® Pan Sharpening

PANCROMA is a powerful new utility for creating true color RGB images and pan sharpened images from Landsat and SPOT^® satellite band files. The program is designed to complement free Landsat databases such as the University of Maryland Global Land Cover Facility (GLCF) database. Landsat data is available from other sources as well, and PANCROMA can handle all of the common formats that you will encounter.

Why is Landsat important? Simply put, Landsat is the only no-cost remotely sensed imagery available for most of the world. So if you are a cartographer, geologist, forester, urban planner, agronomist, climatologist, flight simulator, disaster relief responder, telecommunications planner, water resource manager etc. Landsat is your starting point.

It may seem that services such as Google^® Earth and Bing^® Maps with their global coverage might have obsoleted Landsat. Google^® Earth for example covers much of the earth with high resolution imagery and that coverage is continually improving. There are two reasons why this is not so: license and spectral range. Google^® for example restricts the free use of its data to those applications that can be accessed freely via the internet. So if you are a cartographer making a map of a regional park for example, you cannot simply lift Google^® Map data for your project. Commercial use in effect requires a commercial license. Google is in the advertising business, not the satellite data business and its license terms and conditions favor its interests, not necessarily yours.

Landsat sensors offer a broader range of data products as well. While Google and Bing images span the visible spectrum from about 0.45 microns to 0.70 microns, Landsat has six multispectral sensors that expand this range out to 2.35 microns. Landsat also has a very useful thermal infrared (TIR) band that spans 10.4-12.5 microns. This is important as certain features of the earth's surface absorb/reflect preferentially in the longer wavelength bands. A few examples of Landsat 30m and 15m resolution image samples are shown to the right. (The 30m resolution images are shown as insets in the pan sharpened images.) These additional bands are super-useful for many applications ranging from land use to gold prospecting.

The main problem with Landsat is the relatively low resolution of its band files: 28.5m-30m per image pixel. At one time this was cutting edge for public-domain satellite data. Around 2000 the United States government decided to allow commercialization of the type of satellite imagery that was only previously available to the national intelligence services. As a result, Landsat 30m resolution is decidedly medium resolution today. Pan sharpening can easily double this resolution to 15m per pixel, greatly increasing the value of the imagery. Some samples of Landsat 30m images are shown to the upper right of this page.

Another major problem with Landsat is the 2003 failure of its Scan Line Corrector. This has resulted in all Landsat ETM+ data collected after that time to have striped gaps in the images resulting in a 22% data loss in each image. The application has utilities for filling these gaps with pre-2003 data in order to restore most of the usefulness of this information.

Landsat data files are very large. Each of the three band files needed for pan sharpening is about 60MB in size. The panchromatic file is even larger: about 250MB. Several temporary files of similar size are required to perform the necessary transformations. As a result, pan sharpening uses a lot of computer resources and more powerful computer is preferred for the compute-intensive algorithms required especially for pan sharpening.

The primary purpose of the Landsat survey satellites is to provide scientists with data to assess the earth's vegetation, crop and mineral resources. It was not designed to produce true color images of the earth's surface. However, Landsat bands 3, 2, and 1 are close to the red, green and blue (respectively) wavelengths that the human eye is sensitive too. These bands can be inserted into an RGB color model and will produce images that are very satisfactory for visualization. Scientists often combine other Landsat bands in order to produce false-color images for other purposes. For example, bands 4, 3, 2 are commonly used to highlight vegetation. Many other combinations of bands are used for different purposes.

The latest Landsat sensor is the Enhanced Thematic Mapper Plus (ETM+). Among other improvements, ETM+ included for the first time a 15m resolution panchromatic band. Rather than detecting radiation in a narrow range of frequencies, a panchromatic sensor collects radiated energy across a broader spectrum. And it does so at increased resolution compared to previous sensors, i.e. 15m/pixel as opposed to the 30m/pixel resolution of the band sensors. The panchromatic data is not recorded in separate RGB channels but instead is recorded in a single channel. So despite having greatly increased resolution compared to bands 1-7, the panchromatic band can only be displayed as a black and white image. The corresponding visible RGB color data is stored separately in channels 3, 2, and 1, but at a lower (30m) resolution. Looked at another way, each band channel contains only 25% of the pixels and visual information of the panchromatic channel.

If there were a way to combine the color information in these channels with the higher resolution image information in the panchromatic channel then it would be possible to produce color images at the higher (15m) resolution. This is precisely what the PANCROMA algorithms do: produce 15m color image that can look as natural as the lower resolution images produced by combining the band data. Because Landsat covers virtually the entire earth, you can produce such images for any area that interests you. Unlike image data extracted from sources such as Google Earth™, Landsat is copyright free.

PANCROMA offers several pan sharpening algorithms, including ENHG, AJISANE, XIONG, Brovey, and SPECTRAL. However, with the advent of the new ELIN local optimization algorithm, it is possible for virtually anyone to produce high quality pan sharpened images using the Hue, Saturation, Intensity method in only a few steps. The procedure for doing so is as follows:

To begin the process, Select 'File' | 'Open'. Successively select the band images in the order: blue; green; red. In addition, open one additional file, namely the associated panchromatic image with the …nn80… (GeoTiff) or '…_B80.L1G' suffix as well. It is recommended that you run your first panchromatic computation with a subsetted Landsat image set, prepared as described above. This will save a lot of computing time while you learn the method and iterate the process in order to achieve the best results. PANCROMA can automatically subset your Landsat scene to a matching file set using its subsetting utilities. When all four files are opened successfully, the relevant menu selections and check boxes will be enabled. If you do not wish to set any pre or post image processing parameters, select 'Pan Sharpen' from the menu bar. You will be presented with a choice of several pan sharpening methods, for example: 'HSI Transform' or 'Brovey Transform'. PANCROMA provides several different algorithms for producing panchromatic images. Different algorithms are necessary in order to best compensate for spectral distortions inherent with different satellite data sets. The HSI transform is a good one to start with.

After opening your files and selecting the HSI processing option, you will be presented with a dialog box prompting you to select the number of interpolation iterations. (PANCROMA offers several interpolation options. However the default Nearest Neighbor works very well for Landsat). Accept the default settings on the data entry dialog box. Note that the ELIN local optimization algorithm is enabled as the default.

After a bit of time that will depend on your computer and the size of your data files, your pan sharpened image will appear. You can save this in one of several formats. PNG is optimal for general graphic work. GeoTiff is necessary if you need to retain your metadata, which is often the case. A full pan sharpened Landsat scene is a very large file: some can approach 1GB in size. Landsat files are the largest that most people will ever encounter in any context. An example of a Landsat 30m color composite image, and the corresponding pan sharpened 15m resolution Landsat image is shown to the right.

This tutorial has shown a few of the features of the PANCROMA satellite image processing system. The application has many more useful features that are explained in detail in the PANCROMA Instruction Manual. A free trial version of the software is also available for download as well.