基于图像分割的SAR图像匹配方法

针对单一背景的海岛合成孔径雷达（SAR）图像，提出一种基于图像分割的图像匹配方法．该方法采用静态小波对原实时图分解滤波，抑制噪声；利用最大类间方差阈值法分割图像，分割出特定目标；选择结构元素进行形态腐蚀和膨胀，去掉细小背景杂波干扰；对二值化的图像进行面积统计，搜索出面积最大的目标．在原实时图中，以最大面积目标的外切矩形裁剪图像作为新的实时图．实践证明，利用裁剪的新实时图进行匹配，可以有效去除背景对匹配的不利影响，弥补了传统方法匹配精度低、实时性差的缺点．

SAR-image matching by using image segmentation

From SAR-images of island against single background, an image matching method based on image segmentation is proposed. Five processes were used in the algorithm, a. Static wavelets were employed to eliminate speckle noises;b. The given objects were segmented out the segmented images by Otsu threshold;c. Structural elements were selected and eroded and expanded with morphologic operations to eliminate the interference of background noises; d. Two-valued images were counted by area to search the objects with the maxim area; e. The exrectangle of the objects with maximum area divided up as real-time images in the original real-time image. The tests show that the effects of the background on image matching can be eliminated and the shortcomings in traditional method, low precision and poor real time, improved by this new image segmentation algorithm.