An information theory perspective on computational vision

作     者：Alan YUILLE 

作者机构：Department of StatisticsUniversity of California at Los AngelesLos AngelesCA 90095USA 

出 版 物：《Frontiers of Electrical and Electronic Engineering in China》 (中国电气与电子工程前沿（英文版）)

年 卷 期：2010年第5卷第3期

页      面：329-346页

学科分类：0808[工学-电气工程] 0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 0805[工学-材料科学与工程（可授工学、理学学位）] 0702[理学-物理学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：The author would like to acknowledge funding support from NSF with grants IIS-0917141 and 0613563 and from AFOSR FA9550-08-1-0489 

主　　题：computer vision pattern recognition information theory minimum description length Markov random field(MRF)model stochastic grammars 

摘      要：This paper introduces computer vision from an information theory *** discuss how vision can be thought of as a decoding problem where the goal is to find the most efficient encoding of the visual *** requires probabilistic models which are capable of capturing the complexity and ambiguities of natural *** start by describing classic Markov Random Field(MRF)models of *** stress the importance of having efficient inference and learning algorithms for these models and emphasize those approaches which use concepts from information *** we introduce more powerful image models that have recently been developed and which are better able to deal with the complexities of natural *** models use stochastic grammars and hierarchical *** are trained using images from increasingly large ***,we described how techniques from information theory can be used to analyze vision models and measure the effectiveness of different visual cues.