This is a little more complicated. You could subset the image by a vector or ROI using RESIZE_DOIT. Then create a mask from the vectors/ROI, apply the mask to mask the values outside of the boundary, and then do a WHERE on the remaining pixels that are not masked. I have a sample program that would do the resize and masking that uses some undocumented routines. You would need to add in the code for WHERE that will find the indices of the pixels that meet the criteria 'of a certain value'. ========= pro subset_by_evf compile_opt idl2 ;select input file to subset fname = 'C:\Program Files\ITT\IDLxx\products\envixx\data\world_dem' envi_open_file, fname, r_fid=fid if (fid[0] eq -1) then return envi_file_query, fid, bnames=bnames, sname=sname ;open vector file to use for subset and get id evf_fname = 'C:\Program Files\ITT\IDLxx\products\envixx\data\vector\states_.evf' envi_check_save, /vector evf_open, evf_fname, vec=vec, /no_warning evf_id = vec.id ;compute the spatial boundary based on the evf using ROI routines roi_ids = envi_get_roi_ids(fid=fid, /include_evfs) ; this is an undocumented routine that finds the bondaries of an ROI and populates the DIMS array envi_roi_compute_spatial_boundry, roi_ids, dims, fid=fid ;resize the original image by the new dimensions envi_doit, 'resize_doit', fid=fid, dims=dims, pos=pos, $ rfact=[1.,1.], /in_memory, r_fid=sub_fid envi_file_query, sub_fid, ns=ns, nl=nl, dims=dims ;generate a mask from the evf - undocumented routine to generate a mask envi_mask_doit, evf_id=evf_id, evf_fid=sub_fid, ns=ns, nl=nl, and_or=0, $ /inside, /in_memory, r_fid=m_fid ;apply the mask to the subset image, setting background pixels to -9999 (or whatever you like depending on the data type) out_name=sname+'_subset' envi_mask_apply_doit, fid=sub_fid, pos=pos, dims=dims, m_fid=m_fid, m_pos=0l, $ value=-9999, out_bname=bnames, out_name=out_name ;close in memory files envi_file_mng, id=m_fid, /remove envi_file_mng, id=sub_fid, /remove envi_evf_close, evf_id end