图像分割中空间一致性和无监督聚类算法应用

Space Consistency and Unsupervised Clustering Algorithm Application in Image Segmentation

图像空间结构中的图像一致性和无监督聚类信息在图像分割过程中有着非常重要的作用,提出应用图像空间一致性和无监督聚类算法来达到快速执行图像分割的新算法.首先,利用概率树结构针对目标图像进行图像分割过程中形成的过分割区域,并使得这些区域能够达到理想边缘部分;然后,将用基于概率的无监督图像分割框架来处理分割区域.概率树结构结合了以往传统树结构的相关优点,能够更加自然地针对对象的相关边界进行框架建模.本文提出的算法不需要在每次迭代执行过程中更新相关自适应结构,从而可以大大减少计算所需要的时间.通过非参数密度估计技术结合传统的精确反馈框架,此类评价方法针对人类对象而言具有更为强大的边界不一致.

The structure of the image space image consistency and unsupervised clustering information, has a very important role in the process of image segmentation, this paper presents the application of image space consistency and unsupervised clustering algorithm to achieve the rapid implementation of a new algorithm for image segmentation. First, the use of a probability tree structures formed over divided regions for the target image,the image segmentation process, and makes these areas to be able to achieve the desired edge por- tion; unsupervised image segmentation framework based on the probability of then will be used to process the divided regions. Proba- bility tree structure combines the advantages of the traditional tree structure in the past and more natural framework for modeling object boundary. The proposed algorithm does not need to be updated in each iteration during the execution of the adaptive structures, which can greatly reduce the time required for the calculation. Nonparametric density estimation techniques combined with accurate feedback framework, the type of evaluation method for human subjects with a more powerful boundary inconsistency.