Adaptive striping watershed segmentation method for processing microscopic images of overlapping irregular-shaped and...

Area of Science:

• Image analysis
• Computer vision
• Morphological algorithms

Background:

• The morphological watershed algorithm often suffers from oversegmentation.
• This issue arises not only from irregular particle shapes but also from particles with multiple centers, like ellipses.

Purpose of the Study:

• To investigate the causes of oversegmentation in watershed algorithms.
• To develop a novel method for suppressing oversegmentation in particle image analysis.

Main Methods:

• Introduced a new parameter: the striping level.
• Developed a criterion for the striping parameter to find markers before segmentation.
• Established an adaptive striping watershed algorithm using a marker searching algorithm.

Main Results:

• The proposed method effectively suppresses oversegmentation.
• Demonstrated effectiveness on typical particle images, including gold nanorod ensembles.
• Identified multiple centers as a cause of oversegmentation.

Conclusions:

• The adaptive striping watershed algorithm successfully mitigates oversegmentation.
• The striping level parameter aids in accurate marker identification for improved segmentation.
• This approach enhances the analysis of particle images, particularly those with complex shapes or multiple centers.