PANCROMA contains an implementation of the AJISANE transform. This algorithm is similar in some ways to the HSI algorithm. It is capable of transforming any number of spectral channels into an orthogonal color space qualified by an intensity vector. As a result, it is possible to transform the blue, green, red, and near infrared (NIR) bands into the AJISANE space, substitute the panchromatic band for the intensity band, and then reverse transform back to RGB space.
The advantage of such a four-dimensional transform is that some panchromatic bands (notably Landsat) overlap into the NIR band. As a result, substitution of the panchromatic band for the intensity band in the HSI algorithm can result in spectral distortions. Because the AJISANE space can contain NIR information as well as RGB information, this type of spectral distortion can be reduced. AJISANE is more memory expensive than XIONG, the other PANCROMA five-file pan sharpening method. However, it can offer an alternative for those cases where the use of XIONG results in unwanted artifacts as described in a previous article.
To use the AJISANE transform, open five band files as in XIONG processing by selecting 'File' | 'Open', opening the blue, green, red, panchromatic, and near infrared bands in this exact order. As always, the band files must all be of exactly the same dimensions and the panchromatic band must be a 2X, 3X, 4X or 6X multiple of the band file sizes. When the files are open, select 'Pan Sharpening' | 'AJISANE Transform'. The Pan Sharpening Input box will become visible. You must input the number of interpolation iterations using the drop-down combo box. PANCROMA will automatically select Laplacian interpolation for any panchromatic file size greater than 2X and will assign a default interpolation factor of 10. If you wish to assign more Laplacian iterations, or if you wish to add Thin Plate iterations you must check the 'Activate Image Processing Routines' check box on the Main Window. When you do so, you can check the 'Thin Plate Check Box' and assign Thin Plate iterations. When you are finished, click on the 'Process Image' check box. After some processing time, your pan sharpened image will be computed. If you have left the 'Display Pan Sharpened Image' check box checked, your image will be displayed. This method of pan sharpening, like the Brovey, can result in images that are a bit dark. If this is the case, you can brighten the image by checking the '
Enable Channel Level Adjust' on the Image Processing Data Input box.
An example of a pan sharpened image using this method is shown to the right. In addition to adding ten units of brightness to each channel as described above, I also used the sharpening and contrast adjustment utilities on the Image Processing Data Input box. The color quality is fairly good, as can be seen by comparing the pan sharpened image with the RGB color composite from the three band files (second image to the right).
|
|