SAR Images Change Detection Using Markov Random Field Based On BM3D

Change detection is a technique to locate and analyze multiple remote sensing images obtained from the same geographical location and at different times,so as to obtain and identify the information of the change of the ground objects.Because Synthetic Aperture Radar(SAR)is not sensitive to light and atmospheric conditions,the target can be detected all-weather and all-day,so detection based on changes in SAR images is widely used in military reconnaissance,environmental monitoring,urban change,and disaster assessment.With the continuous development and maturity of SAR imaging technology,change detection is increasingly important in the application of SAR images.Since SAR is a coherent imaging system,it is inevitably affected by the speckle noise,which will make it very difficult for subsequent change detection and analysis,and greatly increase the false detection rate of change detection.Therefore,how to choose the appropriate algorithm to suppress the speckle noise and improve the image detection accuracy is the difficulty and focus of SAR image change detection.The traditional change detection algorithm uses two images obtained at the same time and at different times to obtain the difference image by pixel-by-pixel ratio,and then uses the threshold method to segment the ratio image to complete the change detection of the image.However,the algorithm does not take into account the spatial correlation of image pixels and the speckle noise inherent in SAR images,making the accuracy of change detection lower.The Markov Random Field(MRF)model can make full use of the spatial context information of the image to suppress the speckle noise,so it can greatly improve the accuracy of SAR image change detection and is widely used in SAR image change detection.The difference image of the traditional Markov Random field change detection algorithm is obtained by comparing the pixel-by-pixel ratios of two SAR images of different phases.This method is greatly affected by the speckle noise,and the detection accuracy is low.This paper improves this method.We use the weighted average of the gray values of the eight neighboring pixel points of a single pixel to replace the gray value of the pixel point.Then we obtain the difference image by comparing the two SAR images of different phases on a pixel by pixel basis.Finally,the MRF model is used to segment the difference image,so that the influence of speckle noise can be better suppressed and the accuracy of SAR image change detection can be improved.In order to better suppress the influence of speckle noise,this paper proposes a Markov random field SAR image change detection based on BM3 D.The algorithm firstly obtains the logarithm ratio of two SAR images of different phases on a pixel-by-pixel basis.The ratio image transforms the multiplicative speckle noise into additive noise,and then uses the BM3 D algorithm to filter the log ratio image to obtain the final difference image.Finally,the MRF model is used to segment the difference image to obtain the variation detection result of the SAR image.The BM3 D filter uses the similarity of the image itself to estimate the original gray value of the image pixel point by comparing the similarity between the image blocks at different positions in the image.This method makes full use of the spatial and temporal correlation a nd the speckle noise.It is suppressed well.This paper uses real SAR image data to verify the proposed algorithm.Experimental results verify the effectiveness of the proposed algorithm.