Noisy Remote Sensing Image Segmentation with Wavelet Shrinkage and Graph Cuts
- 255 Downloads
In this paper, a new noisy remote sensing image segmentation algorithm combined with wavelet shrinkage and graph cuts model is proposed. The entire process of noisy remote sensing image segmentation is composed of two steps. Firstly, the wavelet transform is used to extract information about sharp variations in the remote sensing images and the shrinkage function is applied to adapt the image features, and image noise is eliminated by utilizing the feature adaptive threshold method. Secondly, graph cuts based on active contour (GCBAC) model is applied to segment the de-noised image. Additionally, a new energy function which disregards the regularising parameter is proposed in the GCBAC model in order to avoid the edge and region balance problems, and the GCBAC model is used to extract the desired segmentation object by constructing a specified graph. Simulation results indicate that the proposed algorithm can effectively improve the quality of image segmentation and demonstrates improved robustness to noise.
KeywordsNoisy remote sensing image Wavelet shrinkage Graph cuts Feature adaptive Threshold
This work was supported in part by the International Cooperative Research and Personnel Training Projects of the Ministry of Education of the People’s Republic of China [Grant number DICE2014-2029].
- Gupta, K. K., & Gupta, R. (2007). Feature adaptive wavelet shrinkage for image denoising (pp. 81–85). Chenai: IE- EE-ICSCN, MIT Campus, Anna University.Google Scholar
- Juan, O., & Boykov, Y. (2006). Active graph cuts. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1, 1023–1029.Google Scholar