A Study on Robust Image Matching between Images Hard to Correspond and its Application

为了找到难以匹配的两幅图像之间的对应关系，提出了一种内窥镜图像的匹配方法，即包含重复模式的图像间匹配方法。与通常的方法相比，这些方法具有较强的稳健性和准确性。作为这些方法的应用，我们还提出了一种肠道CT图像的扩展方法及其频谱分析，并给出了从两个角度重建形状的代价函数。我们已经证明了这些方法的有效性。

In order to find correspondences between two images hard to match, we have proposed a matching method for endoscopic image of the intestine, matching method between the images including repetitive patterns. These methods are robust and accurate compared with the usual methods. As the applications of these methods, we have also proposed an extension method for the intestine CT image and its frequency analysis and a cost function for a shape reconstructed from two views. We have shown the effectiveness of these methods.