一个在图像中识别变化区域的有效方法

An Efficient Approach to Detect Changed Regions in the Image

本文介绍了一个新颖有效的方法,用于估计图片中的变化区域.本文利用现有的一维参数变点估计方法设计了一个可以应用到图像分割问题中的方法.新方法采用了Schwartz信息量准则来估计变点个数,然后通过一个改进后的PELT算法来计算变点位置.此外,在估计完变点之后,本文也提出一个全新的方法可以将同分布的区域聚合在一起.我们证明了在一些合适的条件下,变点的估计和区域的估计均是相合的.在数值模拟研究中,新方法在估计精度和计算时间等方面都要优于其他的图像分割算法.

In this paper,we propose a novel and efficient method for detecting distribution changes in a picture.We design an approach based on the parametric change-points detection problem and apply it into region detection problem.The number of change-points is determined by the Schwartz information criterion and the location of the change-points is estimated via a modified pruned exact linear time algorithm.Furthermore,we propose a new merging method to merge the regions in the same distribution after finishing the estimation of change-points.Under some mild conditions,we prove our estimation of change-points and regions is consistent.In the simulation studies,the new method performs well in both estimation and computation time compared to other method.