论文部分内容阅读
为了解决腹腔软组织电子计算机断层扫描影像难于分割的难题,提出一种基于格式塔认知框架的多目标分割算法.通过借鉴格式塔认知框架中“邻近度、相似度”的思想,引入超像素算法处理电子计算机断层扫描图像处理,生成可视块.进一步地,在可视块粒度描述有向邻接关系,以软组织的相对空间位置约束聚类分割过程.在公开数据库上的实验结果表明,该算法降低了聚类的计算量,其结果比当前流行的算法准确率更高.
In order to solve the difficult problem of intractable computerized tomography (CT) images, a multi-objective segmentation algorithm based on Gestalt cognitive framework is proposed. By referring to the idea of “proximity, similarity ” in cognitive framework of gestalt, The superpixel algorithm processes the computed tomography image processing to generate the visual block.Furthermore, the directional adjacency relationship is described on the visual block granularity, and the clustering segmentation process is constrained by the relative spatial position of the soft tissue.Experimental results on the public database , This algorithm reduces the computational complexity of clustering, and the result is more accurate than the current popular algorithm.