Clustering Method based on Fusion of Multi-spectral Images

B. Qiu, M. Herbin, and E. Perrier (France)


Multi-spectral, fusion, clustering,segmentation


Image segmentation is a very common problem in image processing field with many applications, which results in different research branches such as change detection, object tracking, segmentation of maps, multi temporal filtering, etc. This paper focuses on clustering problem based on fusion of multi-spectral images/satellite photos, which belongs to the remote sensing field. The steps are as follows: First, the projection from n-D to 1-D is used as the fusion process of multi-spectral images; then based on the fusion result, K-means method is chosen to cluster pixels into different classes, which should represent different regions. For the demand of our project FLOCODS, river regions are paid to high attention. The results of the tests proved our idea feasible and good when compared with another software in this field, Multispec.

Important Links:

Go Back