Add Support for Diamond-Shaped Structuring Element in Morphological Operations #27441
Conversation
|
@KAVYANSHTYAGI, thank you for the contribution! Can you please provide some references to papers, articles where diamond-shaped structuring elements are used? It's the first time I hear about it. In OpenCV you can more or less easily specify structuring element of any form, so may be we don't need extra feature which is just for convenience. We'd like to keep the library compact unless there is some clear advantage of extending it. |
There was a problem hiding this comment.
Thank you for the feedback! I appreciate the emphasis on keeping OpenCV concise and adding features with broad utility.
References
"Morphological Image Analysis: Principles and Applications"
Pierre Soille — Section 3.1.3 covers the diamond (L1-metric) structuring element and its uses.
Springer Link
"Fundamentals of Digital Image Processing"
Anil K. Jain — Discusses diamond shapes as a basic structuring element.
Lam, Lee, and Suen, “Thinning Methodologies—A Comprehensive Survey,” IEEE TPAMI, 1992.
Diamond SE is used in skeletonization/thinning algorithms.
IEEE Link
“Efficient Implementation of Morphological Operations with Arbitrary Structuring Elements,” IEEE, 1994.
Efficient algorithms for diamond SE.
(https://ieeexplore.ieee.org/document/296353)
scikit-image library --- Offers diamond SE as a built-in primitive.
Justification
The diamond structuring element is a standard primitive in mathematical morphology, especially for L1/Manhattan distance operations. It’s widely documented in textbooks and research, and is essential for connectivity based tasks like skeletonization and thinning.
While OpenCV supports custom kernels, having diamond as a built in, well tested option ensures consistency, reduces user error, and matches what’s already standard in libraries like scikit image. This makes foundational morphology algorithms easier to use, reliable, and reproducible across different users and codebases.
Not the correct IEEE link...
Quick Google Scholar returns nothing...
This would deserve a direct link instead of reading a full paper, false url or non-existing paper... |
Pull Request Description
Overview
This PR introduces support for a new diamond-shaped structuring element in OpenCV’s morphology module. The addition enables users to perform morphological operations (e.g., erosion, dilation, opening, closing) using a diamond-shaped kernel, which is defined by Manhattan distance from the center.
Detailed Changes
Motivation
Diamond-shaped structuring elements are widely used in image processing, particularly where neighborhood connectivity is defined by Manhattan distance (e.g., certain denoising, thinning, or skeletonization tasks). By supporting this shape natively, OpenCV expands its flexibility and applicability to new classes of problems.
Backward Compatibility
This change is fully backward compatible and does not alter any existing functionality or APIs. It only adds a new option for kernel shape selection.
Usage Example
cv::Mat diamondKernel = cv::getStructuringElement(cv::MORPH_DIAMOND, cv::Size(5, 5));Testing
Checklist
Added new enum values for the diamond shape.
Implemented diamond kernel logic in morphology dispatch.
Closing
This PR fills a gap in OpenCV’s morphological toolkit by providing a commonly-used structuring element. Please let me know if further changes or documentation are needed.