Share Email Print
cover

Proceedings Paper

A region-based fusion rule using K-means clustering for multi-sensor image fusion
Author(s): Xiao-zhu Xie; Ya-wei Xu
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Region-based fusion rule is adopted in research on multi-sensor image fusion because of its better accuracy than pixel based and widow-based fusion rules. Background subtraction is a simple technique to divided image into object region and background region but its uses are limited because a set of continuous images is needed. Moreover, it’s difficult to distinguish objects from visible image without continuous images when objects are pretended. A region-based fusion rule using K-means clustering is proposed to overcome the limit stated earlier. In the proposed scheme, low frequency images produced by DT-CWT (Dual-Tree Complex Wavelet Transform) were clustered into different regions to obtain a joint map. According to local energy ratio (LER) and the size of regions, average gradient of window or ratio of region sharpness (RS) are adopted as measures respectively to fuse low frequency coefficients of different regions in joint map. The experimental result shows that the proposed approach performs better than pixel-based and window-based fusion rules in entropy, spatial frequency, average frequency and QAB/F.

Paper Details

Date Published: 3 December 2015
PDF: 5 pages
Proc. SPIE 9794, Sixth International Conference on Electronics and Information Engineering, 97941V (3 December 2015); doi: 10.1117/12.2201136
Show Author Affiliations
Xiao-zhu Xie, Academy of Armored Force Engineering (China)
Ya-wei Xu, Academy of Armored Force Engineering (China)


Published in SPIE Proceedings Vol. 9794:
Sixth International Conference on Electronics and Information Engineering
Qiang Zhang, Editor(s)

© SPIE. Terms of Use
Back to Top