A segmentation algorithm for noisy images
This paper presents a 2-D image segmentation algorithm and addresses issues related to its performance on noisy images. The algorithm segments an image by first constructing a minimum spanning tree representation of the image and then partitioning the spanning tree into sub-trees representing different homogeneous regions. The spanning tree is partitioned in such a way that the sum of gray-level variations over all partitioned subtrees is minimized under the constraints that each subtree has at least a specified number of pixels and two adjacent subtrees have significantly different ``average`` gray-levels. Two types of noise, transmission errors and Gaussian additive noise. are considered and their effects on the segmentation algorithm are studied. Evaluation results have shown that the segmentation algorithm is robust in the presence of these two types of noise.
- Research Organization:
- Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
- Sponsoring Organization:
- USDOE Office of Energy Research, Washington, DC (United States)
- DOE Contract Number:
- AC05-96OR22464
- OSTI ID:
- 437768
- Report Number(s):
- CONF-9611123-1; ON: DE96014695
- Resource Relation:
- Conference: IEEE symposium on image, speech, natural language systems, Washington, DC (United States), 4-6 Nov 1996; Other Information: PBD: [1996]
- Country of Publication:
- United States
- Language:
- English
Similar Records
2-D image segmentation using minimum spanning trees
Patch-based image segmentation of satellite imagery using minimum spanning tree construction