Abstract: Image segmentation methods based on graph cuts can achieve a global optimal solution in binary labeling problem, and get a local optimal solution with strong features in multiple-labels problem. But it has a bad effect on the complex images that have noise or occlusions, so it can not make people satisfy. This paper proposes an image segmentation method based on shape prior segmentation and graph cut. On the basis of graph cut, incorporated with shape prior knowledge, the algorithm can restrict the object search space by more constraint, extract the target completely and improve the algorithm accuracy, and handle the affine transformation of shape by feature matching algorithm, in order to make this algorithm flexibility and deal with different situations. The results of expermient show that the proposed method is effective.

Key words: image segmentation, graph cut, shape prior, energy function

摘要： 基于图割理论的图像分割方法在二值标号问题中可以获取全局最优解，而在多标号问题中可以获取带有很强特征的局部最优解。但对于含有噪声或遮挡物等复杂的图像，分割结果不完整，效果并不令人满意，提出了一种基于形状先验和图割的图像分割方法。以图割算法为基础，加入形状先验知识，使该算法包含更多约束信息，从而限制感兴趣区域的搜寻空间，能够更好地分割出完整的目标，增加了算法的精确度。针对形状的仿射变换，运用特征匹配算法进行处理，使算法更加具有灵活性，能够应对不同类型的情况。实验表明了该算法的有效性。

关键词: 图像分割, 图割, 形状先验, 能量函数