Abstract
A new segmentation method based on mean shift and fuzzy integral algorithm is proposed to overcome the problems of hyperspectral image processing. A hyperspectral image is grouped into several band subset images according to their correlation relationship. Then, the dimension of each subset band image is reduced by principle component analysis. To segment each subset band image quickly, the mean shift algorithm is employed to find the cluster centers. And, the segmentation results are fused by fuzzy integral. Simulation results show the effectiveness of this method.
| Original language | English |
|---|---|
| Pages (from-to) | 188-192 |
| Number of pages | 5 |
| Journal | Guangzi Xuebao/Acta Photonica Sinica |
| Volume | 39 |
| Issue number | 1 |
| DOIs | |
| State | Published - Jan 2010 |
Keywords
- Feature dimension reduction
- Fuzzy integral
- Hyperspectral image
- Image segmentation
- Mean shift