bwhitmiss on a sparse set of locations in MATLAB

Say I have a huge image I, and a set of pixel locations indexed by pix_locations (linear indices indexing locations in I). I would like to run a hit-and-miss operation (i.e. what bwhitmiss does) with a structuring elements SE only on the locations inde开发者_如何学运维xed by pix_locations.

I know I can just run bwhitmiss on the entire image, and then keep the results corresponding to the locations indexed by pix_locations, but if the image is large, this is a waste and slow (the locations indexed by pix_locations are sparse).

Are there any ways to run bwhitmiss selectively on a set of locations on an image?

I suppose you could partition your image, like a K-d tree, than use the bwhitmiss on the non-empty partitions, but I think as bwhitmiss is a simple look-up table operation you wouldn't benefit much from that micro-optimization unless you have really huge and sparse images.