In what way was creating the mask from the ROIs not successful? This is a very common way to create a mask.

Maybe i have expressed myself false about "succeful". i went this way: ENVI-Classifikation-Neural Net- Select Input File- "Maximum_likelihood classified image..(for example)-Mask Options-Build Mask- Select Attributes for Mask- Import ROIs( i click for Import ROIs, i think its common way)- then i have choosed from my 6 Klasses 3 of them to mask them out and click OK, but i have then error message : "Neural net classification need at least 2 bands for classifier". And this is my problem. I want to take my classified image (from max like), mask some crop classes and let them flow in Neural Net to classify them. Additionally i just wanna ask for right masking. You mean, it is right to import ROIs for already classified image to perform masking operations? Thank you in advance!!!

You can't classify an already classified image. I was thinking you needed to create a mask from areas of your first classification to apply to the multispectral image during your second classification so that it would ignore some areas. You can make a mask of a variety of sources - ROIs, shapefiles, data ranges, etc. But you can't classify an already classified image since there is no spectral information available in a classified image.

Ooook, but if so, maybe it was what i have to do, because i actually have a schema and have understood it like depicted above (classify classified image). But what you are saying, it sounds like this is appropriate way for me: " I was thinking you needed to create a mask from areas of your first classification to apply to the multispectral image during your second classification so that it would ignore some areas " And in this case, what i have to do? create my mask from already classified image and apply to the image for second classification? thanks so much!!