SciPy 2014 sprint ideas

• Clean-up
  • Move and deprecate functions
    • canny from filter to feature
    • threshold_* from filter to exposure
  • Refactor canny detector so that we can use non-maximal filtering on its own (this would have been useful for teaching)
  • Add default structuring elements to morphology operations, and fall back on scipy.ndimage when parameters don't fit our optimized functions based on rank filters (2D images, bitdepth<12).
• Fixes
  • Fix watershed tie-breaking for boundaries (Issue 803)
  • Update active image for skimage.viewer tools when a new image is opened (the displayed image changes, but the image used for filter doesn't currently change) (See mailing list)
  • Fix blitting behavior for skimage.viewer. (PR #1030)
• Improvements
  • Add int64 support to our img_as_* functions
  • Deprecate filter module and rename to filters or filtering (so we don't shadow the builtin filter)
  • Distribute tifffile.py with scikit-image and make it the default for TIFF files
  • Other IO improvements, such as reordering plugin preferences, issuing warnings when certain plugins are used that rescale the image to float, etc.
• Features
  • Add IPython widgets for replicating skimage.viewer.canvastools
  • Add flag to normalize gradients in Canny edge detector so that the threshold values don't need to be adjusted as you change sigma
  • There's also an old list of requested features
  • Improve 3D support (a few attendees requested this and expressed interest in helping with that)