Skeletonize command behavior

[Davide Heller]

Dear all,

     I am using the default Skeletonize command (Process>Binary>Skeletonize) to process a cell tissue segmentation result (see example https://www.dropbox.com/sh/xeg2cgh35qr9g6g/5sVHvWRbZ1 skeletonize_input.tif, 4kb). While 1.46m (1.46r tested) produces a connected set of lines in location [30,40] (skeletonize_result_1.46m.tif, 4kb), the future versions (first tested 1.47h, last 1.48h, skeletonize_result_1.47h-48h.tif, 4kb) do produce an unconnected version. I noticed a change to the Skeletonize command introduced with version 1.47d, but I am not sure whether this is the cause.

Assuming 8-connectivity of the skeletonization, actually also the previous result has a flaw preserving a 4px-square in the same location. Could anybody comment on which behavior is the correct one or if both are wrong.

Many thanks,
Davide

******************************************

Davide Heller
PhD candidate
Mering & Basler laboratory
Institute of Molecular Life Sciences

[phone]

Room Y55-L-48
University of Zurich
Winterthurerstrasse 190
CH-8057

******************************************


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Rasband, Wayne (NIH/NIMH) [E]]

On Dec 2, 2013, at 1:12 PM, Davide Heller wrote:

> Dear all,
>
>     I am using the default Skeletonize command (Process>Binary>Skeletonize) to process a cell tissue segmentation result (see example https://www.dropbox.com/sh/xeg2cgh35qr9g6g/5sVHvWRbZ1 skeletonize_input.tif, 4kb). While 1.46m (1.46r tested) produces a connected set of lines in location [30,40] (skeletonize_result_1.46m.tif, 4kb), the future versions (first tested 1.47h, last 1.48h, skeletonize_result_1.47h-48h.tif, 4kb) do produce an unconnected version. I noticed a change to the Skeletonize command introduced with version 1.47d, but I am not sure whether this is the cause.
>
> Assuming 8-connectivity of the skeletonization, actually also the previous result has a flaw preserving a 4px-square in the same location. Could anybody comment on which behavior is the correct one or if both are wrong.

Both behaviors are incorrect and both are fixed in the latest ImageJ daily build (1.48i2).

-wayne
--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html