Statistical Multisensor Image Segmentation in Complex Wavelet Domains
We propose an automated image segmentation algorithm for segmenting multisensor images, in which the texture features are extracted based on the wavelet transform and modeled by generalized Gaussian distribution (GGD). First, the image is roughly segmented into textured and non-textured regions in the dual-tree complex wavelet transform (DT-CWT) domain. A multiscale segmentation is then applied to the resulting regions according to the local texture characteristics. Finally, a novel statistical region merging algorithm is introduced by measuring a Kullback-Leibler distance (KLD) between estimated GGD models for the neighboring segments. Experiments demonstrate that our algorithm achieves superior segmentation results.
Keywordsmultisensor image segmentation statistical modeling complex wavelets
- Lewis, J., Nikolov, S., Canagarajah, N., Bull, D., Toet, A.: Uni-Modal Versus Joint Segmentation for Region-Based Image Fusion. In: Proc. of the Int. Conf. Information Fusion, pp. 1–8 (2006)Google Scholar
- Nixon, M., Aguado, A.: Feature Extraction and Image Processing. Academic Press, Oxford (2008)Google Scholar
- Simoncelli, E., Anderson, E.: Noise Removal via Bayesian Wavelet Coring. In: Proc. of the IEEE Int. Conf. Image Process., pp. 378–382 (1996)Google Scholar
- Wan, T., Canagarajah, N., Achim, A.: Multiscale Color-Texture Image Segmentation with Adaptive Region Merging. In: Proc. of the IEEE Int. Conf. Acoustics, Speech, and Signal Process., pp. 1213–1216 (2007)Google Scholar
- Wan, T., Canagarajah, N., Achim, A.: Statistical Multiscale Image Segmentation via Alpha-Stable Modeling. In: Proc. of the IEEE Int. Conf. Image Process., pp. 357–360 (2007)Google Scholar