Statistical properties of natural 3D scenes and their role in visual perception

还有什么能比看到更容易呢？我们不需要尝试就能看到，而且通常也不需要思考。事实上，看东西是一项了不起的壮举。这似乎很简单，因为我们大脑中负责视觉的部分是一台功能强大的计算机，具有一生的看东西经验。视觉具有挑战性的一个原因是，所有的视觉图像都是高度模糊的，可以想象，任何给定的图像都可以被视为描绘了广泛的形状、颜色和光线。我们的视觉系统需要以某种方式克服这种模糊性，以便正确地感知场景和对象。我们的视觉系统克服模棱两可的一种方式是依赖于对最有可能出现的形状、颜色和光线的假设。例如，之前的实验表明，从婴儿期开始，我们的视觉系统就会慢慢学习到，光线通常从我们上方发出，我们的视觉系统利用这一知识正确地感知物体和场景。本研究项目提出的主要问题是，人类视觉系统为了正确感知世界，还依赖于哪些关于真实世界形状、颜色和照明的其他事实？我们将通过两种方式调查这个问题。首先，使用新的测量工具，我们将测量世界的三个感知重要特征：表面形状、表面颜色和照度。我们将对这些测量进行统计分析，以了解视觉系统可以使用真实世界的形状、颜色和光线具有哪些可靠的特性，以便正确地感知场景和对象。其次，利用知觉实验，我们将调查关于形状、表面颜色和照明的假设如何指导人类对场景和对象的感知。观察者将判断精心设计的计算机生成的模糊对象的图像的形状、颜色和照明，并使用计算和数学模型来推断视觉系统依赖于哪些假设来从这些模糊的图像中正确地看到场景和对象。这项研究将有助于我们理解人类正常和异常的视觉，也将有助于设计正确地看到场景和物体的计算机视觉系统。

What could be easier than seeing? We see without trying, and usually without thinking about it. In fact, seeing is a remarkable feat. It seems easy because the part of our brain responsible for vision is a powerful computer with a lifetime of experience at seeing things. One reason why vision is challenging is that all visual images are highly ambiguous, and any given image could conceivably be seen as depicting a wide range of shapes, colours, and lights. Our visual system needs to overcome this ambiguity somehow, in order to perceive scenes and objects correctly. One way our visual system overcomes ambiguity is by relying on assumptions about what shapes, colours, and lights are most likely to occur. For example, previous experiments have shown that starting from infancy, our visual system slowly learns that light usually shines from above us, and our visual system uses this knowledge to perceive objects and scenes correctly. The main question addressed by the present research project is, what other facts about real-world shapes, colours, and lighting does the human visual system rely on in order to perceive the world correctly? We will investigate this question in two ways. First, using novel measurement tools, we will measure three perceptually important characteristics of the world: surface shape, surface colour, and illumination. We will do statistical analyses of these measurements, in order to see what reliable properties real-world shapes, colours, and lights have that the visual system can use in order to perceive scenes and objects correctly. Second, using perceptual experiments we will investigate what assumptions about shape, surface colour, and illumination guide human perception of scenes and objects. Observers will judge the shapes, colours, and lighting of carefully designed, computer-generated images of ambiguous objects, and using computational and mathematical models we will deduce what assumptions the visual system relies on to see scenes and objects correctly from these ambiguous images. This research will help us understand normal and abnormal human vision, and it will also be useful for designing computer vision systems that see scenes and objects correctly.