Related Experiment Video
Updated: Jul 1, 2026
07:45
Quantifying Intermembrane Distances with Serial Image Dilations
Published on: September 28, 2018
Adaptive pseudo dilation for gestalt edge grouping and contour detection.
Giuseppe Papari1, Nicolai Petkov
1Institute of Mathematics and Computing Science, University of Groningen, Groningen, The Netherlands. g.papari@rug.nl
Summary
This study introduces adaptive pseudo-dilation (APD) for robust object contour detection in natural images. The novel method effectively identifies contours amidst texture and low contrast, improving image analysis.
Area of Science:
- Computer Vision
- Image Processing
- Computational Photography
Background:
- Local edge detection methods struggle with textured regions where luminance changes are high.
- Object contours are often obscured by texture, necessitating global analysis for accurate detection.
- Existing contour detection algorithms may fail with low-contrast boundaries or complex image textures.
Purpose of the Study:
- To develop a novel contour detection method for natural images that overcomes limitations of local edge detectors.
- To introduce an adaptive pseudo-dilation (APD) operator for identifying curvilinear structures in edge maps.
- To enhance contour detection accuracy and robustness against texture and low-contrast features.
Main Methods:
- Introduced adaptive pseudo-dilation (APD), a morphological operator using context-dependent structuring elements.
- Limited dilation to the Voronoi cell of each edge pixel for precise contour localization.
- Developed a multithreshold contour detector embedding an edge pixel grouping algorithm based on APD output.
- Utilized generalized reconstruction from markers for contour completion at multiple threshold levels.
Main Results:
- The APD operator effectively identifies long curvilinear structures, aligning with the Gestalt law of good continuation.
- Grouping edge pixels via APD output demonstrated superior agreement with perceptual grouping principles.
- The multithreshold detector significantly suppressed texture and improved detection of low-contrast contours.
- Qualitative and quantitative comparisons confirmed the proposed method's superiority over existing approaches.
Conclusions:
- The proposed adaptive pseudo-dilation operator offers a novel and effective approach for object contour detection.
- The integrated multithreshold contour detection framework provides robust performance across various image conditions.
- This method advances image analysis by enabling more accurate and reliable identification of object boundaries in complex natural scenes.
