User Manual - pancroma

The default method is checked. The Nearest Neighbor method is a fast andsimple interpolation method that can work well in many cases. However it canresult in ragged edges in areas of abrupt color change.An iterative Laplacian method can alternatively be used. It is the default methodfor data sets where the panchromatic band is more than 2X the multispectralbands, for example pan sharpening Landsat with the SPOT panchromatic bandwhere the ratio is 3X. The Laplacian is relatively fast and it converges reliably.When you check this radio button, a text box will become visible so that you canspecify the number of iterations. The larger the number, the more likely theinterpolation will converge. However this is at the expense of greater processingtime.A problem with the Laplacian however is that it is not very efficient at extractinginformation from beyond its nearest neighbor pixels. This can start to affect theRGB image color quality for 4X data like WorldView2. To remedy this,PANCROMA TM can run a Thin Plate preprocessing interpolation algorithm if youinstruct it to. The Thin Plate algorithm is also fast and it reaches out beyondnearest neighbor pixels to interpolate. One disadvantage is that it can be noisy.That is why it may work better as a preprocessing step using relatively fewiterations than as a primary interpolation method.In order to select the Laplacian (or any other than the default Nearest Neighbormethod) check the ‘Activate Image Processing Routines’ check box on the mainwindow. When the image processing data input box becomes available, selectthe ‘Laplacian’ radio button in the ‘Interpolation Method’ data group. Once this isselected, two ‘Laplacian’ combo boxes and a ‘Thin Plate’ combo box will becomevisible. Select either standard Laplacian or modified Laplacian (Laplacian2).The Laplacian2 interpolation is a modified Laplacian algorithm that reaches outfurther beyond the target pixel in performing its interpolation. It improves pansharpening results for difficult 4X and 6X interpolations. You can add the desirednumber of Laplacian iterations fro the drop down menu. To add Thin Plateiterations, check the ‘Thin Plate Check Box’ and select your desired iterationsfrom its drop down combo box. These will be added to the Laplacian iterationsthat you specified in the pan sharpening data input box. The location of theInterpolation Method data group is shown in the figure below.Some 4X and 6X interpolations can be troublesome. The problem is that neitherthe Laplacian nor Thin Plate algorithms can be absolutely relied upon to producesmooth interpolations. In an interpolation algorithm, the source image pixels aretypically frozen and not modified by the interpolation in order to preserve the dataintegrity the relatively few known values. Rather than interpolating smoothly, theknown pixels may remain standing somewhat proud of the surroundinginterpolated pixels. This can occur under certain arrangements of source imagepixel values. The problem is shown below in an example interpolated hue file.53