[scikit-image] region analysis 3d

[Juan Nunez-Iglesias]

Hi Yan,

For point 1, I think “perimeter” is the wrong word for 3D, but you could add “surface” and make that @only3d. I think this would be a very valuable addition! There’s several variants of this that would be interesting, for example, the number of pixels minus the number of pixels after erosion could approximate the surface area.

I’m not sure I understand point 2. What are you trying to say here? Doing a region-props version where the input is e.g. pixel contours? This is an interesting idea but I would hesitate to overload regionprops, and would instead define a new function, e.g. “coordprops” or “contourprops”. What do you think?

Juan.

On 13 Apr 2018, 9:18 PM +1000, imagepy at sina.com, wrote:
> Hi every one:
>     I want to do some region analysis for 2d/3d block. I saw the skimage.measure.regionprops is very powerful. But if we can add some feature below:
>
>     1. I want to got every 3d block's surface area.
> The regionprops's area for 3d object meas volumn and the perimeter is only2d, I think the method skimage to cound perimeter works for 3d surface, binary_outline it then convolve with a 3*3*3 weight matrix.
> Another function is march_cubes, it returns a series of face and normal vector, but if we can return a index or a lable array, to indicate which faces are in a same block?
>
>     2. I see the regionprops is based on pixels analysis, and in some feature, wo must to copy a new image, I think we can do some work by vector. We can count many feature by contours (except the one need intensity_image).
>
> yxdragon
> _______________________________________________
> scikit-image mailing list
> scikit-image at python.org
> https://mail.python.org/mailman/listinfo/scikit-image
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scikit-image/attachments/20180415/920addb9/attachment.html>