Geodesic DT with occlusion points propagation 2D and 3D

• Algorithm: Ruben Cardenes
• Implementation: Ruben Cardenes
• Date: Oct 2002
• Paper: Occlusion Points Propagation Geodesic Distance Transformation, ICIP 2003 https://www.researchgate.net/publication/221120209_Occlusion_points_propagation_geodesic_distance_transformation

Description

This code implements a geodesic distance map calculation. It is geodesic in the sense that it is constrained to a domain, so distances between source point and any point are not Euclidean if the domain is not convex. See 2D examples.

Installation

Clone the repository. Change to the created directory. Execute "make"

2D version:

This version is purely C code.

Usage:

geodesicDT2D [options] sourcefile.txt domain.png out.png
              -d (debug mode, outputs the oclussion points in png format)
              -c color output (0: gray, 1: red-blue, 2: random, 3: green scale)

• source_points: txt file with coordinates of source points
• input_domain_image: input domain image in png format, positions with zero values are the one where the distance map will be computed, non-zero values are outside the domain
• ouput_image: output distance map image in png format

The 2D version uses png for input and output.

Examples for one point:

geodesicDT2d -c 0 example_data/source_2D_02.txt example_data/domain256_1.png output01.png
geodesicDT2d -c 1 example_data/source_2D_03.txt example_data/domain256_2.png output02.png
geodesicDT2d -c 2 example_data/source_2D_01.txt example_data/domain256_3.png output03.png
geodesicDT2d -c 3 example_data/source_2D_01.txt example_data/domain256_4.png output04.png
geodesicDT2d -c 0 example_data/source_2D_01.txt example_data/domain512_3.png output05.png

Example for several points (4 points)

geodesicDT2d -c 2 example_data/source_2D_4_points.txt example_data/domain256_1.png output06.png

Python wrapper

from geodesicDT2d import run_geodesic2dDT_wrapper

debug = 0
color_mode = 0
run_geodesic2dDT_wrapper('example_data/source_2D_02.txt','example_data/domain256_1.png','output01.png', color_mode, debug)
color_mode = 1
run_geodesic2dDT_wrapper('example_data/source_2D_03.txt','example_data/domain256_2.png','output02.png', color_mode, debug)

3D version

This version is implemented in C and C++

Usage:

geodesicDT3d [options] source_points.txt domain_image.mhf output_image.mhd

• source_points: txt file with coordinates of source points
• domain_image: input image in metaheader data format
• ouput_image: output image in metaheader data format

This 3D version uses input and output format in metaheader format, which is just a text header file and a raw data file as in ITK.org.

Example With opened domain (11 source points):

geodesicDT3d example_data/source_3D_01.txt example_data/domain3d_80_01.mhd out.mhd

Example With closed domain (9 source points):

geodesicDT3d example_data/source_3D_02.txt example_data/domain3d_80_02.mhd out.mhd

Notes

For both versions the input and output are a list of single points defined in a txt file. The code can be easily modified to accept an arbitrary set of pixels given as an image. The domain region of interest are always pixels equal to zero

License

This code is released under the MIT license.

If you use this work for academic or commercial use, please cite the paper: Occlusion Points Propagation Geodesic Distance Transformation, ICIP 2003 https://www.researchgate.net/publication/221120209_Occlusion_points_propagation_geodesic_distance_transformation