Simple python-based interactive image annotation tool requiring only numpy and matplotlib

Dependencies:

matplotlib
numpy
scipy (for loading .mat files, as necessary)

Tested using:

OSX Mavericks, Yosemite
matplotlib 1.4+
numpy 1.10+
scipy 0.14+

Usage:

python image_annotate.py [options] [image_file]
  -h  --help         print this usage and exit
  -r  --ref          reference image (RGB or RGBA image to display, useful if [image_file] contains >4 channels)
  -b  --bands        indices of reference bands (default=[0,1,2])
  -w  --wvl          wavelength file
  -l  --load-rois    load ROIs from file
  -c  --color        color file (default=cm_jet16.txt)
  -n  --norm         L2 normalize each pixel in image_file
  -s  --scale        scaling factor for reference image colors (default=1.0)
  -f  --flip         flip data up/down w.r.t. reference file
  -v  --verbose      enable verbose output

Mouse commands:

  left click:    define new ROI point, drag for free-hand drawing, 
                 select/deselect existing ROI
                 
  right click:   close current ROI polygon, delete existing ROI

Keyboard commands:

  e: export rois
  m: export mean image
  a: annotate ROIs
  d: delete last ROI
  g: group selected ROIs
  u: ungroup selected ROIs
  j: join unlabeled ROIs to labeled ROI group
  c: clear all unlabeled ROIs
  =: grow zoom window 
  -: shrink zoom window
  q: quit (and export if ROIs changed since startup)

Loading matlab (.mat) files as images:

Because scipy.io.loadmat returns a dictonary, it is necessary to provide the key to that dictionary where the image data is located. To do so, use the following syntax

python image_annotate.py [options] /path/to/image.mat:key_value

This syntax also works for the ref and wvl files. For instance, the command:

python image_annotate.py -r /path/to/ref.mat:ref_img -w /path/to/wvl.mat:wavelengths /path/to/image.mat:img_dat

will load the ref_img variable from ref.mat as the reference image, the wavelengths variable from the wvl.mat file as the image wavelengths, and the img_dat variable from image.mat as the image data file.

Credit

Based on code by Daniel Kornhauser from the matplotlib-users mailing list available at the following url: http://www.mail-archive.com/matplotlib-users@lists.sourceforge.net/msg00662.html

dsmbgu8/image_annotate.py