1

PANSHARPEN Error

cboscari

December 19, 2013 10:35

None

Follow

I am (trying to) use Landsat 8 imagery to get recent data for my area of interest. I need to use PANSHARPEN to create higher resolution imagery of bands 4 and 5, using band 8 for the panchromatic image. I am trying this with the PCA option but if fails with any PANSHARPEN option selected.  I am adding band 4 and 5 as my files in the selector, band 8 as my panchromatic image. I receive the error

“Input Image Dimensions do not match.”

 

I have verified that the image Column and Row sizes are identical for bands 4 and 5, but of course are different for band 8. In fact, they are identical except for the range of values in each image. I have also tried using WINDOW to create smaller images, but this did not fix the error. I have also used STRETCH to rescale the images to make the the values consistent, but that did not work either.

 

Any help on what is going on? I know what the error is saying, but it seems my images are the same size (except for the higher res band 8), so I am confused. PANSHARPEN has very little in the way of documentation.

Additional- L7 imagery is too corrupted to use at my area of interest. Hopefully there is a way to get this to work with L8?

• Facebook
• Twitter
• LinkedIn

5 comments

• 0
  
  Clark Labs December 19, 2013 18:51

  Look at the Min and Max X and Y. You should notice that the pan coordinates do not exactly match the MS bands, even though the rows and columns are exactly doubled. I believe you will notice that it may be off by half a pixel. If these do not match, PCA will not run. 

  Couple of things you could do. If you go back to the metadata for the bands, see if IDRISI documented these coordinates correctly. Idrisi uses the lower left corner for its cell origin, not cell center. If it is mis-documneted, you can simply change these values in Metadata in Idrisi Explorer. Otherwise, you can try a resample if the error is larger to force the pan band to match the coordinates of the MS bands. 

  James Toledano

  Clark Labs

• 0
  
  cboscari December 19, 2013 20:12

  Hi, thank you for the quick response.

  You were correct, the the band 8 pan image had a different value for max x and y. I used RESAMPLE on band 8, , and now the Max and Min of  X and Y of all the files match. However, PANSHARPEN *stil*l gives me the error.

  I'm starting to panic.

  Chris

• 0
  
  Clark Labs December 19, 2013 22:47

  Chris,

  What about the rows and columns? The resampled pan band should be exactly double in each from the MS bands. I assume when you ran resample you just used corner points for the gcp's with zero rms? A quicker way to do this is to use Project, which calls Resample. It avoids the creation of the gcp's. 

   

  James

• 0
  
  cboscari December 19, 2013 22:57

  The Rows and Columns are both 1 byte short of double in the original file. The resampled file kept those dimensions but changed the Min and Max X and Y values. I will try project and see what happens.

  Chris

• 0
  
  cboscari December 19, 2013 23:25

  Ah, ha! That seemed to fix it. I told project to increase the rows and columns by one byte, and use the min x, min y, max x and max y from the other bands and now it reads it in. Thanks!

  Chris