基于自适应分块马尔可夫随机场(MRF)模型的SAR图像变化检测研究

The Research of SAR Image Change Detection Based on Adaptive Block MRF Model

现有的合成孔径雷达(SAR)图像变化检测算法仅仅考虑图像的单一像素信息,未考虑图像像素间的空间依赖关系,检测结果易受图像噪声的影响,检测精度较低。将马尔可夫随机场模型引入到SAR图像变化检测算法中,极大地提高了检测精度,但是算法计算量很大,效率极低。基于分块马尔可夫随机场模型的SAR图像变化检测算法极大地提高了检测效率,但是检测过程中采用固定尺寸的图像分块方法,未考虑图像的局部细节特征,检测效果不是很理想。首先根据图像局部细节特征将图像划分成尺寸变化的子图像块,然后利用基于马尔可夫随机场模型的SAR图像变化检测算法实现变化检测。实验结果表明,提出的算法既保持了很高的检测精度,又极大地提高了检测效率。

At present, the algorithms of SAR image change detection only focus on information of one pixel, without considering the spatial correction within the neighbouring pixels of image, so the detection results are sensi- tive to noise and the detection accuracy is very low. The change detection algorithm based on Markov Random Field model can get high accuracy detection results, but the calculation amount is very large and the detection efficiency is very low. The detection efficiency of algorithm based on block MRF model is greatly improved, but the algorithm divides image into invariable size blocks disregarding the image minutiae, so the detection effect is not ideal. The image was divided into variable size blocks based on image minutiae, then by using the algorithm of MRF model, the change detection was implemented. The results of experiments show that the algorithm proposed not only can keep high detection accuracy but also improve the detection efficiency greatly.