|Please use this identifier to cite or link to this publication: http://hdl.handle.net/1926/1292|
The 3D method suggested here is a direct implementation of the work of Lee et al.  who have demonstrated that their algorithm can find all deletable surface points at every iteration and is thus very fast. The code was tested on six clinical datasets.
The autor implements a algorithm for thinning in binary images based on decision trees that works with 3D images.Hypothesis:
Source code and a testing image are included.Open Science:
Sources are included, including a test image and its result, source of latex article and references to the original paper that explains the method.Reproducibility:
The code was downloaded and compiled without issuesQuality of the data :
It's easy to reproduce and generate new skeletons.Interest:
Skeletons are widely used in several fields of computer visions.
ITK as a whole, could benefit of this addition to their libraries as it currently only 2D skeletons are supportedFree comment :
The metod is fast and robust (could not find a volume that made the algorithm to crash).
This submission is related to the family of Binary thinning algorithms to ultimately extract the centerlines of the objects in an image. This is a 3D implementation that follows the decision tree based approach suggested by Ta-Chin Lee et al. The algorithm is fast and accurate. In addition to that it is very easy to use: No user-defined paramters are required except the input and output image type.
In 3D this algorithm can correctly iteratively handle all possible binary combinations of object and background voxels in a 26-neighbourhood and find all deletable surface points at each iteration.Evidence:
The author explained the basis of the algorithms and provided the code along with "how-to-use-it" description and test images. Thus the user is able to test the code on the provided images and compare the results with ones provided by the author.Open Science:
This work follows the Open Science approach. The author does provide the source code that immediately can be tested on accompanying images.Reproducibility:
Compilation of the downloaded code was easy and did not make any troubles. After running the code on the provided images, the results were identical to the ones provided by the author. The accompanying paper/user manual was quite comprehensive and handy.Quality of the data :
Quality of the provided data was good and very easy to use.Interest:
This is a very good (fast and accurate) 3D skeltonazer which can be used in the object tracking software and/or as a building block in a morphological software.Free comment :
A very nice and handy toolbox for the object tracking applications. I would recommend including these classes from this submission to the ITK library.
|Categories:||CMake, Code memory optimization, Code speed optimization, DART, Filtering, Generic Programming, Iterators, Mathematical Morphology, Parallelization, SMP, Programming, Surface extraction|
|Keywords:||3D thinning, skeletonization|
Linked Publications more
Reader/Writer for Analyze Object Maps for ITK
by Hawley J., Johnson H.
Implementation of weighted Dijkstra’s shortest-path algorithm for n-D images
by Weizman L., Freiman M., Joskowicz L.
Send a message to the author