Chen Pengyun, Zhang Yichen, Jia Zhenhong, Yang Jie, Kasabov Nikola
College of Information Science and Engineering, Xinjiang University, Urumuqi 830046, China.
Institute of Image Processing and Pattern Recognition, Shanghai Jiao Tong University, Shanghai 200400, China.
Sensors (Basel). 2017 Jun 6;17(6):1295. doi: 10.3390/s17061295.
Traditional image change detection based on a non-subsampled contourlet transform always ignores the neighborhood information's relationship to the non-subsampled contourlet coefficients, and the detection results are susceptible to noise interference. To address these disadvantages, we propose a denoising method based on the non-subsampled contourlet transform domain that uses the Hidden Markov Tree model (NSCT-HMT) for change detection of remote sensing images. First, the ENVI software is used to calibrate the original remote sensing images. After that, the mean-ratio operation is adopted to obtain the difference image that will be denoised by the NSCT-HMT model. Then, using the Fuzzy Local Information C-means (FLICM) algorithm, the difference image is divided into the change area and unchanged area. The proposed algorithm is applied to a real remote sensing data set. The application results show that the proposed algorithm can effectively suppress clutter noise, and retain more detailed information from the original images. The proposed algorithm has higher detection accuracy than the Markov Random Field-Fuzzy C-means (MRF-FCM), the non-subsampled contourlet transform-Fuzzy C-means clustering (NSCT-FCM), the pointwise approach and graph theory (PA-GT), and the Principal Component Analysis-Nonlocal Means (PCA-NLM) denosing algorithm. Finally, the five algorithms are used to detect the southern boundary of the Gurbantunggut Desert in Xinjiang Uygur Autonomous Region of China, and the results show that the proposed algorithm has the best effect on real remote sensing image change detection.
基于非下采样轮廓波变换的传统图像变化检测总是忽略邻域信息与非下采样轮廓波系数之间的关系，并且检测结果易受噪声干扰。为了解决这些缺点，我们提出了一种基于非下采样轮廓波变换域的去噪方法，该方法使用隐马尔可夫树模型（NSCT-HMT）进行遥感图像变化检测。首先，使用ENVI软件对原始遥感图像进行校准。之后，采用均值比运算得到差异图像，该差异图像将由NSCT-HMT模型进行去噪。然后，使用模糊局部信息C均值（FLICM）算法将差异图像划分为变化区域和未变化区域。所提出的算法应用于真实的遥感数据集。应用结果表明，所提出的算法能够有效抑制杂波噪声，并保留原始图像中更多的细节信息。所提出的算法比马尔可夫随机场-模糊C均值（MRF-FCM）、非下采样轮廓波变换-模糊C均值聚类（NSCT-FCM）、逐点方法和图论（PA-GT）以及主成分分析-非局部均值（PCA-NLM）去噪算法具有更高的检测精度。最后，使用这五种算法对中国新疆维吾尔自治区古尔班通古特沙漠的南边界进行检测，结果表明所提出的算法在真实遥感图像变化检测中效果最佳。
