I'm working with multi-temporal datasets from modis and had to do various vegetation indices. I ran into a problem now because there is a significant amount of water too, which skews the resulting map. How can I build a mask to just blend out the water bodies?


It's been some months since I used ENVI and unfortunately I don't have it available to poke around in right now, but I can recall three ways you could create the mask. The first would be to use the draw tools in ENVI to create shapes to use as a mask; or more appropriately the ROI (region of interest) tool. The second would be to use GIS software to generate the same and save it as a shapefile, or perhaps locate some water body shape information for your area of interest. The third would be to run a classification on the image for the purpose of identifying water pixels and then saving the results to a shapefile that could then be used as a mask.

The following links may also help you with specific steps or alternate methods: http://www.exelisvis.com/docs/ManageRasterData.html#Masking http://hyspeedblog.wordpress.com/2013/10/18/application-tips-for-envi-5-0-building-a-mask-from-a-classified-image/

You may also wish to edit your question to clarify a couple of things. What exactly do you mean by 'skews the resulting map' - too much color ramp devoted to water? When you say 'blend out' you actually mean cut out, correct? Which version of ENVI are you using? That may influence methods available to you.


Transform your dataset in MNF. Use 2D scatter plot using MNF components 1 and 2.

Find the water from the scatter plot:

sample 2D scatter plot

sample mask

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.