How can i convert images from scikit-image to opencv2 and other libraries?

Question

I tried to find contour with cv2 python library in a skeletonized image created with scikit-image and i got this error:

    contours, hierarchy = cv2.findContours(skel,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
TypeError: <unknown> data type = 0 is not supported

My question is: What i have to do to convert to cv2 and viceversa?

I know that opencv use numpy.uint8 type to represent binary images instead scikit-image numpy.float64

I used also mahotas (numpy.bool) and pymorph libraries. How can i convert from scikit-image to these libraries and viceversa?

I think i found myself solution! If i use: skel = np.array(skel, dtype=np.uint8) i easily convert in opencv2 type and so on for other libraries! — improc, Apr 22 '13 at 23:05
http://scikit-image.org/docs/dev/user_guide/data_types.html#using-an-image-from-skimage-with-opencv — tidy, Jul 06 '17 at 06:48

Stefan van der Walt · Answer 1 · 2017-07-20T16:58:35.543

19

scikit-image provides conversion routines between the different data-types that also correctly preserves scaling:

from skimage import img_as_ubyte

cv_image = img_as_ubyte(any_skimage_image)

Update: the scikit-image user guide now has a more detailed section on this: http://scikit-image.org/docs/stable/user_guide/data_types.html#working-with-opencv

edited Jul 20 '17 at 16:58

answered Apr 23 '13 at 11:22

Stefan van der Walt

7,165
1
32
41

3

What would be vice versa variant: from cv_image to skimage? – Yuriy Chernyshov Feb 11 '17 at 18:06
4

If cv_image is an array of unsigned bytes, skimage will understand it by default. If you prefer working with floating point images, `img_as_float` should do the trick. See the skimage user guide for more information. – Stefan van der Walt Feb 11 '17 at 19:40
2

Also, whoever will need these conversions, keep in mind that OpenCV use BGR color space by default while Skimage use RGB one :-) – Yuriy Chernyshov Feb 12 '17 at 00:14
That's a good reminder, thanks Yuriy. I'll file an issue on the scikit-image docs to add a section on this conversion. (https://github.com/scikit-image/scikit-image/issues/2500) – Stefan van der Walt Feb 12 '17 at 20:04
Answer update above with link to new section in documentation. – Stefan van der Walt Aug 25 '17 at 17:29

score 0 · Answer 2 · edited Nov 25 '18 at 09:43

0

This is working for me:

image = skimage.io.imread(mpath)
img = image[:, :, ::-1] # convert image from RGB (skimage) to BGR (opencv)
cv2.namedWindow('original', cv2.WINDOW_NORMAL)
cv2.imshow('original',img)

edited Nov 25 '18 at 09:43

JJJ

32,902
20
89
102

answered Nov 25 '18 at 09:19

oferb

127
1
5

How can i convert images from scikit-image to opencv2 and other libraries?

2 Answers2