基于轮廓线统计量的前景分割Markov随机场模型  

Figure-ground Separation by Contour Statistics and Markov Random Field Model

作　　者：汤慧旋[1] 危辉[1] 

机构地区：[1]复旦大学计算机科学与工程系认知算法模型实验室,上海200433

出　　处：《自动化学报》2009年第8期1033-1040,共8页Acta Automatica Sinica

基　　金：国家重点基础研究发展计划(973计划)(2010CB327900);国家自然科学基金(60303007);上海市科委重点科技攻关项目(08511501703)资助~~

摘　　要：提出了一种基于轮廓线统计量的前景分割Markov随机场(Markov random field,MRF)模型,和Grabcut等以往模型不同,本文模型通过在分割标签的编码中加入对轮廓线方向的考虑,将Gestalt知觉组织的原则加入分割约束中去,从而使分割边界更为平滑,作为前景分割和Gestalt知觉组织原则研究的基本框架,本文模型的系统结构分为前景分割、注意力选择和信息整合三个子模块,与相关神经生理研究的结论相一致,最后,分别给出了基于本文模型的自动和半自动前景分割实现,结果好于Grabcut等相关算法的结果。In this paper, we propose a Markov random field （MRF） based representation for the Gestalt law, and suggest using a message passing-like scheme to infer the segmentation. Different from other grabcut models, our MRF function is specially encoded to consider orientations along the contour, thus the Gestalt law is embedded into the inference. As a basic framework of the research in figure-ground separation and Gestalt law, our system is designed in reference to neurophysiology, and the architecture is composed of three modules： primal visual cortex （V1）, extra-striate cortex （V2）, and the interest selected region. To validate our method, we conduct experiments in both auto and interactive segmentation algorithm. The results are better than those of grabcut and other related algorithms.

关 键 词：神经视觉 前景分割 轮廓线 自然统计量 格式塔原则 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]