«

Alasdair McAndrew <amc...@gmail.com> on Thu, 29 Nov 2012 wrote:
> Probably the combinations of OpenCV, Scipy.ndimage and scikits-image
> would cover pretty much all of my needs.

Hi,

All of those (+ mahotas, which is the package I wrote & imread which
might be useful for microscopy file formats) will work on numpy arrays
(openCV requires a bit of conversion back and forth, but it will work).
Therefore, it is not much of a bother to mix&match functions from one or
the other library::

import mahotas as mah
import imread
import skimage
import pylab

image = imread.imread("my-fancy-image.jpeg")
filtered = mah.gaussian_filter(image, 4.)
segmented = skimage.segmentation.quickshift(filtered)
pylab.imshow(segmented)
....

I just mixed 4 different packages seamlessly in this example. This will
work flawlessly.

mahotas & skimage are both under very active development, but neither is
unstable (i.e., we keep adding new features, but the existing code is
very reliable). scipy.ndimage is sort of dead: I scavenged its code for
good bits and pieces for mahotas, but nobody is developing ndimage.

If you run into specific issues, the mailing list pythonvision at

https://groups.google.com/forum/#!forum/pythonvision

is a good forum. Plenty of people from different projects lurk there.

HTH,
Luis