针对局部目标与背景的低对比度、目标边缘模糊红外手印的分割问题, 提出了一种基于结构形态几何生长的边缘模糊红外目标提取算法。该算法首先利用最大熵法及阈值扩展将图像进行粗分割; 在粗分割的目标区域中提取区域块特征点, 构筑手的结构形态; 通过区域块特征点寻找种子点, 并以种子点与对应特征点的距离关系作为生长判决条件进行几何生长, 最终提取得到目标图像。与常用提取算法进行比对实验, 结果表明, 针对边缘模糊的红外手印图像, 本文算法能更有效地提取出完整目标。

To solve the problems of low contrast between a local target and its background and segmentation of an infrared handprint image with blurred edge, an edge blurred infrared target extraction algorithm based on geometric growth with structural morphology is proposed. Firstly, the maximum entropy method and threshold extension are used to segment the image roughly. Secondly, the feature points of the regional blocks in the roughly segmented target area are extracted to construct the structural morphology of the hand. Finally, by using the regional block feature point to find the seed point and using the distance relationship between the seed points and the corresponding feature points as the growth decision condition for geometric growth, the target area is obtained from the image. The method is compared with the conventional extraction algorithm experimentally. The results show that the proposed method can extract the complete target more effectively for the blurred infrared images with blurred edge.