An algorithm for noisy image segmentation
This paper presents a segmentation algorithm for gray-level images and addresses issues related to its performance on noisy images. It formulates an image segmentation problem as a partition of an image into (arbitrarily-shaped) connected regions to minimize the sum of gray-level variations over all partitioned regions, under the constraints that (1) each partitioned region has at least a specified number of pixels, and (2) two adjacent regions have significantly different {open_quotes}average{close_quotes} gray-levels. To overcome the computational difficulty of directly solving this problem, a minimum spanning tree representation of a gray-level image has been developed. With this tree representation, an image segmentation problem is effectively reduced to a tree partitioning problem, which can be solved efficiently. To evaluate the algorithm, the authors have studied how noise affects the performance of the algorithm. Two types of noise, transmission noise and Gaussian additive noise, are considered, and their effects on both phases of the algorithm, construction of a tree representation and partition of a tree, are studied. Evaluation results have shown that the algorithm is stable and robust in the presence of these types of noise.
- Research Organization:
- Oak Ridge National Lab., TN (United States)
- Sponsoring Organization:
- USDOE Office of Energy Research, Washington, DC (United States)
- DOE Contract Number:
- AC05-84OR21400
- OSTI ID:
- 522710
- Report Number(s):
- CONF-9705121--4; ON: DE97008474
- Country of Publication:
- United States
- Language:
- English
Similar Records
2-D image segmentation using minimum spanning trees
SAR Imagery Segmentation by Statistical Region Growing and Hierarchical Merging