Normal 0 false false false MicrosoftInternetExplorer4 Hi ENVI community, I'm having a problem with masking out features of a texture filter analysis. I'm hoping someone on here can help me out. Thanks in advance! Problem: My area of interest (military base) is heavily fragmented with roads. I want to filter for texture (both first and second order in various window sizes) a satellite image over this area of interest. However, I do not want to include any reflectance data from the roads in my analysis. That is, I'm only interested in capturing texture from pixels which represent vegetated areas. I want to carry out this analysis in ENVI because the program has the capabilities of using multiple 2nd order texture filters which is important to my analysis. It'd be great to somehow label the roads pixels as "no data" across my area of interest. However, as I understand it, ENVI does not recognize "no data" as a legitimate pixel value. Therefore, I am trying to create a mask to exclude roads from my analysis. I have tried to go about this approach two ways which I detail below. Both give me conflicting results. Method 1: ArcGIS approach a) Create a vector shapefile mask in ArcGIS which excludes all roads. b) Apply the mask to my landsat scene (bands 1-5, 7, and NDVI) in ArcGIS to come up with an Arc Raster file of my landsat spectral bands that only includes vegetated regions at my study area (that is with roads masked from the scene) that I want processed for texture. The pixels that were masked now read "no data". Therefore, they should be excluded from texture filtering. d) I then save this as a TIFF file, then import into ENVI and process for texture. Problem with this approach: As far as I can tell, ENVI doesn't know what to do with "no data" pixels. So, the program automatically labels these "no data" pixels as 255. You can manually tell ENVI how to read these pixels - but, I believe it has to be a numerical value. This influences the output from the texture processing since this information of pixels labeled 255 is included in most of my moving window sizes (1x1, 3x3, 5x5, 7x7). I have many roads at my study area close to sampling locations that have been masked. Therefore, this is a big problem for my analysis.  Method 2: ENVI approach a) I used my vector shapefile masks from Method 1 step a). and brought this into ENVI along with my landsat bands. b) I then set up the texture filter in ENVI to only process the spatial subset of my vector shapefile for a particular band (either of 1-5, 7, and/or NDVI). c) It somewhat works. Yet as far as I can tell, ENVI still seems to be processing quite a bit of information from areas that are supposed to be masked. The way I'm checking to see if this is working correctly is by processing first-order mean in a 3x3 window using the "Focal Stats" option in ArcGIS. This analysis correctly excludes pixels valued as "no data". I then do my Method 2: ENVI approach. After comparing this versus the ArcGIS approach I notice the same pixels, in respective image outputs obtained from the two differing methods, have quite different values. I believe there is a way to get around this. But, I must have some flaws in my approaches? Would you recommend modifications, or new methodology? I'm a bit stumped and would greatly appreciate any help/advice anyone could offer. Thanks for your time! Eric