Vision as Bayesian inference: analysis by synthesis?

@article{citeulike:828644, abstract = {We argue that the study of human vision should be aimed at determining how humans perform natural tasks with natural images. Attempts to understand the phenomenology of vision from artificial stimuli, although worthwhile as a starting point, can lead to faulty generalizations about visual systems, because of the enormous complexity of natural images. Dealing with this complexity is daunting, but Bayesian inference on structured probability distributions offers the ability to design theories of vision that can deal with the complexity of natural images, and that use `analysis by synthesis' strategies with intriguing similarities to the brain. We examine these strategies using recent examples from computer vision, and outline some important imlications for cognitive science.}, author = {Yuille, Alan and Kersten, Daniel }, booktitle = {Probabilistic models of cognition}, citeulike-article-id = {828644}, comment = {自然形态的场景高度复杂，而很多关于人类视觉的研究往往使用简单的人造的刺激（最著名的莫过于Hubel\&Wiesel最早关于初级视觉皮层的实验）。作者认为处理复杂的自然场景需要贝叶斯推理关于结构化的概率分布。通过综合来分析，我们的大脑处理信息的过程显然是分层次的逐步抽象，大脑的综合能力惊人，在一个不熟悉的场景我们总是能很快找出其主要特征，形成对它的认识。何为结构化的概率分布？}, doi = {http://dx.doi.org/10.1016/j.tics.2006.05.002}, journal = {Trends in Cognitive Sciences}, keywords = {bayesian, decision, inference, making}, month = {July}, number = {7}, pages = {301--308}, posted-at = {2007-11-04 03:21:16}, priority = {1}, title = {Vision as Bayesian inference: analysis by synthesis?}, url = {http://dx.doi.org/10.1016/j.tics.2006.05.002}, volume = {10}, year = {2006} }

TY - JOUR ID - citeulike:828644 L3 - citeulike-article-id:828644 N2 - We argue that the study of human vision should be aimed at determining how humans perform natural tasks with natural images. Attempts to understand the phenomenology of vision from artificial stimuli, although worthwhile as a starting point, can lead to faulty generalizations about visual systems, because of the enormous complexity of natural images. Dealing with this complexity is daunting, but Bayesian inference on structured probability distributions offers the ability to design theories of vision that can deal with the complexity of natural images, and that use `analysis by synthesis' strategies with intriguing similarities to the brain. We examine these strategies using recent examples from computer vision, and outline some important imlications for cognitive science. IS - 7 JF - Trends in Cognitive Sciences EP - 308 TI - Vision as Bayesian inference: analysis by synthesis? VL - 10 T2 - Probabilistic models of cognition SP - 301 KW - bayesian KW - decision KW - inference KW - making N1 - 自然形态的场景高度复杂，而很多关于人类视觉的研究往往使用简单的人造的刺激（最著名的莫过于Hubel&Wiesel最早关于初级视觉皮层的实验）。作者认为处理复杂的自然场景需要贝叶斯推理关于结构化的概率分布。通过综合来分析，我们的大脑处理信息的过程显然是分层次的逐步抽象，大脑的综合能力惊人，在一个不熟悉的场景我们总是能很快找出其主要特征，形成对它的认识。何为结构化的概率分布？ AU - Yuille, Alan AU - Kersten, Daniel PY - 2006/07// UR - http://dx.doi.org/10.1016/j.tics.2006.05.002 ER -