SPEAR Vertical Stripe Removal - NV5 Geospatial

Prev Go to previous topic

Go to next topic

Last Post 27 Feb 2017 09:03 AM by anon

SPEAR Vertical Stripe Removal

3 Replies

Sort:


You are not authorized to post a reply.

Author

Messages

anon

New Member

Posts:51

27 Feb 2017 09:03 AM
Hello! I have hyperspectral images from an airborne sensor that I des-stripe using the SPEAR vertical stripe removal tool. For these images it is most effective when there are no pixels included in the mask, so I set those parameters to 0% on the bottom and 100% at the top of the range. However, with some of my images, not all, I end up with bands that have no data - i.e. all pixels are NaN for those bands, even when the original raster has valid data in the same bands. What is happening here? I can't find any information on what the SPEAR tool actually does to destripe the image. I have a general overview on how it works, but more information would be very helpful, so it's less of a black box. And if anyone has an idea of why I get bands of NaN after destriping, that would be great! I can't continue my processing with those bands. Thank you!

MariM

Veteran Member

Posts:2396
Veteran Member

01 Mar 2017 10:13 AM
What is the result if you allow the mask to use the 5% to 95% pixels? The tool is taking the mean of each pixel column and normalizing it to the band mean. So my guess is that you have some bad pixels in some of the bands of the dataset, which isn't uncommon with hyperspectral data.

Deleted User

New Member

Posts:51

01 Mar 2017 12:04 PM
Hello! Same result when mask is left at default 5%, 95%. I have a total of 24 pixels of NaN in this image of 51,522,240 pixels, so I could understand if the NaN pixel happens to be in that band, then it will make the mean NaN and then the destriped band will be NaN. However, when I take a subset of the raster, and isolate just the bands that ended up NaN in the previous destriping, and destripe just those bands, then they do not all end up as NaN. To be specific, when the entire raster is destriped (170 bands total) bands 52-57, 59-60, & 62-63 are NaN. When just bands 51-64 are destriped, bands 53,54,56,57,60,62,63 are NaN (NOT 52,55,59). That makes no sense to me, if each band is destriped independently from the others. Why would it be different? Does that make the problem a little more clear? Thank you!

MariM

Veteran Member

Posts:2396
Veteran Member

02 Mar 2017 08:52 AM
Does it help if you mask the NaNs and set them to 0? I agree that the destripe should happen per band so it does not make sense that spectrally subsetting out the bands that have NaN should affect the result.

You are not authorized to post a reply.