I am guessing that the problem is that the SVM classification doesn't allow you to specify a mask during processing. So, perhaps you will need to choose a different classification method, at least for your second classification. - Peg Exelis VIS Tech Support

No,no, my problem is generely that i cannot mask out areas ( not the ROIS)for whole satellite image. So if i import ROIs to define the mask for certain class on the image, i have then mask only for ROI. Ok, with classification, that is not only SVM, i used neural net, not SVM. If i first of all mask out certain class from ROI (for example water), then i have new mask band in available band list- but only from ROI-area. But OK, then i tried to take my satellite image and to add this mask band.Then in parameter-window from Neural Net (for example)- i have also to choose classes from the list. I tried both: selected all available ROI-classes from List AND only WATER (for example). But both of them did not work, if i let run the classifier. I have only RMS-window and then directly file in available list. If i open this file, i have nothing but ROI from water on the mask band, but i want to have all water areas on the whole scene! I do not understand...what is the problem...

I don't really understand your explanation very well. But it kind of sounds like you are making a mask from one of your input ROIs to the first classification. And perhaps what you really want to do is create a mask of all of the areas that are eventually assigned to the class corresponding to that ROI, and use that mask to leave those areas out of your second classification. Is that right? In that case, you could create your mask based on a data range (not an ROI) from the first classification result. The min and max values would be whatever pixel value is assigned to that class. Then specify that mask band when you pick your input file for the second classification. - Peg Exelis VIS Tech Support

Yees, it sounds well with this option for data range! But i have tried to mask out with this option and i have taken min and max values from Postclassification-class statistics option. In the created file for statistics for my class ( for example sugar beet) i have 5 bands (rapid eye has 5 bands) with min and max values in each spectral band "Basic Stats Min Max Mean Stdev Band 1 6108 9965 7744.977252 382.875189 Band 2 5799 10423 7831.060298 542.042851 Band 3 5694 10193 7776.664980 665.491912 Band 4 5525 9710 7359.127939 593.264478 Band 5 5824 10690 7588.239311 557.132041" I think,this is what i need as pixel value for data range in my masking? But it doesnt work if i take these values in "build mask" from classification file. I have either black or white mask band....i tried to use pixel values from min value to max value of one band (one from 5 spectral bands) and also min value from 1 band to max value from last 5 band.It does not work...what am i doing wrong? Thanks a lot for your very kindly help!!

You can create a mask as I described from the classification output image. Then you can use that mask on the original image when you start your second classification. I don't really follow what you are doing, but it sounds like you will end up masking out your whole image. If you take the min and max values for a whole band of the input data, and create a mask based on that, it should cover the whole image, right? So, that's probably not going to be helpful for you. - Peg Exelis VIS

In the class statistics for classified image file i have used for min and max - number of class. So for example, sugar beet has in class statistics band number 6 in min and 6 also in max, because i think, the classificated file has only class numbers, like 1, 2, 3 or 4..and so on. I hope its right?