A Study Of SAR Image Registration Method Based On Sparse Structure And SIFT Feature

Image registration is the process of overlaying two or more images of the same scene taken at different times, from different viewpoints, and/or by different sensors. It is a critical technology in image processing and analysis. Image registration has been widely applied to remote sensing data analysis, computer vision, medical image analysis,geographical detection, pattern recognition and so on. Synthetic aperture radar(SAR)systems are all-weather, all-time, long distance, high resolution and has strongly penetrating ability, thus the technique has been widely applied in many fields of national security and economy. The presence of speckle noise in SAR image hampered the following SAR images interpretation, so the existing registration algorithms which are successfully used in optical remote sensing image are usually not applicable to SAR image directly. On the basis of analyzing and summarizing the current image registration techniques and the feature of SAR image, this paper presents a registration method for SAR image can obtain structure area which contain rich structural information. Then SIFT feature points can be extracted from the area. Through a series of experiments, we can prove the feasibility of the algorithm, the completed works is as follows:1. We proposed an image registration method based on block matching and SIFT feature extraction.The presence of speckle lead to the degradation of gray distribution and objects’ properties, and makes a badly undesired effect on feature extraction and feature matching. Based on the block-matching method which using the distance of PPB,proposed method can obtain the multi-direction structural area. SIFT feature points can extracted from the multi-direction structural area with high stability and rich structural information. Thus, it weaken the interference caused by speckle effectively, meanwhile,the rate of false matching reduce rapidly.2. We proposed a fast registration method for SAR image based on spatial sparse structure and SIFT feature. The algorithm makes full use of thining structure and multi-direction structure feature, it can obtain structure sparsity of each pixel. Then,feature points can extracted for the sparse structure area of SAR image and matched effectively. Besides, the outlier matching can be removed with the Random SampleConsensus Criterion(RANSAC). Thus, it reduce the rate of false matching effectively,meanwhile, registration efficiency improved rapidly.