|
|
|
| Wavelet Markov Random Field Based on Context and Hidden Class Label for SAR Image Segmentation |
| Zhang Qiang①;Wu Yan①② |
| ①School of Electronic Engineering, Xidian Univ., Xi’an 710071, China; ②National Key Lab. of Radar Signal Processing , Xidian Univ., Xi’an 710071, China |
|
|
|
|
Abstract Because of the property that Synthetic Aperture Radar (SAR) images include plenty of multiplicative speckle noise, an effective image segmentation algorithm is proposed based on the wavelet hidden-class-label Markov Random Field (MRF) to suppress the affect of speckle. To consider the clustering and persistence of wavelet, the hidden-class-label MRF is extended to the wavelet domain with a new wavelet model for segmented image named hidden-class-label mixture heavy-tailed model, and interscale transition probability is estimated with improved context, then a new Maximum A Posteriori (MAP) classification is obtained. The experimental results show that the method suppresses the affect of noise effectively to achieve exact and robust segmentation result.
|
|
Received: 19 June 2006
|
|
|
|
|
|
|
|
|
2008, Vol. 30 
