PCI Geomatics Help Center

How can we help you today?

Pansharpening in Geomatica Banff

PCI Geomatics -

Geomatica's pan sharpening algorithm fuses the high-resolution panchromatic and low-resolution multi-spectral imagery together to create an enhanced high-resolution color image. The high-resolution color image preserves the original color fidelity and allows for better visualization and interpretation.

Panchromatic data can be fused with multi-spectral imagery acquired simultaneously by the same sensor, or images from different sensors can be used. However, the best results will be achieved when the imagery is collected simultaneously and the resolutions of the panchromatic and multi-spectral data are closely matched. The spectral characteristics of the original data will be preserved in the resulting high resolution color imagery. This means that analysis such as classification can be done on the pan- sharpened imagery with the added benefit of higher spatial resolution.

PANSHARP is an add-on feature that requires the Geomatica Prime and Pansharpening modules to operate and it is available from EASI, Focus (Algorithm Librarian) or the Modeler environment. 

Setup

  1. In the Focus menu bar, click on the Tools dropdown menu and select Algorithm Librarian…
  2. In the Algorithm Librarian panel, click on the Find… button
  3. In the Find what: input text box, type pansharp
  4. Click Find Next 
  5. With the PANSHARP algorithm selected, click on the Open… button in Algorithm Librarian to open the PANSHARP Module Control Panel

mceclip0.png

In the Files tab:

  1. Expand the Input: Input Multispectral Image Channels branch in the Input Ports section
  2. Expand the associated Files branch and the associated multispectral image file so that the image channels are available. 
    • About Input Multispectral ChannelsThis setup step simply tells the algorithm which channels the pansharpening operation should be applied to. The output will only include the channels that you specify in this step.
  3. Check mark the channels you want to input
  4. Collapse the Input: Input Mutlispectral Image Channels branch
  5. Expand the InputRef: Reference Image Channels branch
  6. Make sure that the associated Files and multispectral image file branches are expanded so that the image channels are available. 
    • About Reference ChannelsThe reference channels tell the algorithm which multispectral channels fall within the same portion of the spectrum that the panchromatic image covers. For example, the Pleiades PAN band registers incoming energy between the wavelengths of 480nm and 830nm. The Blue and NIR bands do not completely fall within this spectral range, but they fall within the panchromatic band’s range enough that they should be included as reference layers. Users should not include MS bands that are completely or mostly outside of the portion of the spectrum that the panchromatic band covers.  
  7. Check mark the channels you want to use
  8. Collapse the InputRef: Reference Image Channels branch
  9. Expand the InputPan: Panchromatic Image channel
  10. Make sure that the associated Files and panchromatic image file branches are expanded so that the single panchromatic image channel is available
  11. Check the only panchromatic channel
  12. In the Output Ports section, check the Viewer-RGB options

mceclip1.png

The following table is a list of the reference bands for some well-known satellite sensors:

Sensor Reference Bands
Landsat 7 (ETM+)   Green: 2        Red: 3  Near IR: 4 
SPOT 1, 2, 3 (HRV)   Green: 1 Red: 2  
SPOT 5 (HRG)
Blue: 1 

Green: 2

   
IRS 1C, 1D

 

  Green: 1   Red: 2  
IKONOS Blue: 1   Green: 2 Red: 3   Near IR: 4
QuickBird
Blue: 1 Green: 2  Red: 3  Near IR: 4

In the Input Params 1 tab:

  1. Specify the Source background option
    • File metadata: Reads the NoData value from the input file's NO_DATA_VALUE metadata. The file-level metadata tag is used as the default for every channel; channel-level tags, when available, override the default. When metadata does not exist, every pixel in the input is considered valid.
    • 0: every pixel, in every channel, having a value of 0 is considered NoData
    • None: every pixel, in every channel, is considered valid
    • Specify values: every pixel, in every channel, having the value(s) specified in the Source Background Values parameter is considered NoData
    • File metadata, else specify values: Same as File Metadata, but uses the specified Source Background Values to define the NoData value for every channel, when metadata does not exist.
  2. Specify the Source background value if necessary.
    • Specify multiple values in a comma-delimited list. The first value is applied to the first channel, the second value to the second channel, and so on. If fewer values are specified than the number of input channels, the last value is repeated for all remaining channels. If more values are specified than the number of input channels, the extra values are ignored.
      For example:
      • File metadata, 0: uses file-level metadata. When metadata is unavailable, pixels that have a value of 0 are considered NoData.
      • 255: every pixel, in every channel, that has a value of 255 is considered NoData
  3. Ensure that Enhance Pansharpening is on (YES)
  4. Resample Method
    • Bilinear interpolation: Bilinear interpolation determines the gray level from the weighted average of the nine closest pixels to the specified input coordinates and assigns that value to the output coordinates. This method generates an image with a smoother appearance than nearest neighbor, but the gray-level values are altered in the process, which can result in blurring or loss of image resolution. Similar to cubic-convolution resampling, interpolation is most appropriate for continuous data. Bilinear interpolation is recommended if further analysis (such as classification) using the output is expected, as it will keep the output values closer to the original MS pixel values.
    • Cubic convolution: Cubic convolution determines the gray level from the weighted average of the 16 pixels closest to the specified input coordinates and assigns that value to the output coordinates. The resulting image is slightly sharper than one produced with bilinear interpolation, and it does not have the disjointed appearance produced by nearest neighbor. Similar to bilinear interpolation, cubic convolution is most appropriate for continuous data.
  5. Click Run

mceclip2.png

The image below on the left shows the Pleiades_PansharpProduct (product from Astrium). The image below to the right shows the melbourne_PSH layer (product produced by PCI Geomatica)

Have more questions? Submit a request

Comments

Powered by Zendesk