A robust method for picture segmentation based on a split-and-merge procedure
Abstract
The split-and-merge (SM) algorithm of Horowitz and Pavlidis (1974) is modified and applied to improve the stability of image segmentation in the analysis of satellite remote-sensing data. The basic principles of the SM method and the modifications introduced are described; a sample image is analyzed; a new graph representation is developed to increase the robustness of the algorithm with respect to geometric and photometric disturbances (thus improving stability); the extension of the graph representation to temporal survey sequences is explained; and a sample application tracing the evolution of the thermal structure of an agricultural region on the basis of NOAA-7 AVHRR images (Cheevasuvit et al., 1985) is presented. The applicability of the modified SM approach to segmentation problems in IR imaging, biomedical imaging, or X-ray quality control is indicated.
- Publication:
-
Computer Vision Graphics and Image Processing
- Pub Date:
- June 1986
- Bibcode:
- 1986CVGIP..34..268C
- Keywords:
-
- Image Analysis;
- Image Processing;
- Pattern Recognition;
- Satellite Imagery;
- Temporal Resolution;
- Advanced Very High Resolution Radiometer;
- Algorithms;
- Graph Theory;
- Remote Sensing;
- Robustness (Mathematics);
- Thermal Mapping;
- Instrumentation and Photography