Description
Unfortunately, there is no global agreement between developers of different image processing tools on the preferable coordinate convention yet. I'm aware of 3 most commonly used ones: cartesian (x - horiz. to the right, y - vert. to the top), xy (x - horiz. to the right, y - vert. to the bottom), rc (r - vert. to the bottom, c - horiz. to the right).
Even current skimage algorithms work with two of the above. Things become harder when people start to use several libraries in parallel (i.e. skimage + OpenCV).
I feel pretty sad reimplementing these routines over and over again for the new projects, and will be really happy to have them in the skimage distribution.
Above said, I propose to add to skimage.util: cart_to_xy, cart_to_rc, xy_to_rc and corresponding inverse functions, which will operate on list of coordinates and, possibly, images to convert points/image between these coordinate systems.
Please, let me know how you feel about this.
Description
Unfortunately, there is no global agreement between developers of different image processing tools on the preferable coordinate convention yet. I'm aware of 3 most commonly used ones:
cartesian(x - horiz. to the right, y - vert. to the top),xy(x - horiz. to the right, y - vert. to the bottom),rc(r - vert. to the bottom, c - horiz. to the right).Even current
skimagealgorithms work with two of the above. Things become harder when people start to use several libraries in parallel (i.e.skimage+OpenCV).I feel pretty sad reimplementing these routines over and over again for the new projects, and will be really happy to have them in the
skimagedistribution.Above said, I propose to add to
skimage.util:cart_to_xy,cart_to_rc,xy_to_rcand corresponding inverse functions, which will operate on list of coordinates and, possibly, images to convert points/image between these coordinate systems.Please, let me know how you feel about this.