贝叶斯概率图像自动分割研究

A Study on Bayesian Probabilistic Image Automatic Segmentation

探讨了一种新的图像自动分割的方法。提出应用高斯有限混合模型与期望-极大化算法对图像特征空间的数据进行聚类,采用信息理论准则(ITC)确定要分割的图像区域数目,用贝叶斯概率分割图像。整合这些技术可以实现图像自动分割,而且实验结果表明信息理论准则可以确定适当的区域数目。

A novel method for automatic image segmentation is invesligated. It is proposed that to apply a Gaussian finite mixture model with EM algorithm to cluster the data in image feature space, to adopt the information theoretical criteria (ITC) to determine how many regions should be segmented on a given image, to apply Bayesian probabilistic classification for the segmentation of the given image. Combining these techniques can implement automatic image segmentation, and experiments show that reasonable region number can be determined by ITC.