Visual coding of natural scene statistics

自然场景统计的视觉编码

• 批准号: 341032-2009
• 负责人: Johnson, Aaron
• 金额: $ 1.09万
• 依托单位: Concordia University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2009
• 资助国家: 加拿大
• 起止时间: 2009-01-01 至 2010-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=429510
• 关键词: Visual; coding; natural; scene; statistics

The environment that humans live in is complex, yet our brains must perform visual tasks within a few hundred milliseconds in order for us to survive. How we do this is a fundamental question in psychology. Recent research has shown that our environment has shaped the design of our perceptual systems through evolution and experience. Therefore, direct measurement of the statistical regularities in natural environments (scenes) has great potential value for advancing our understanding of visual perception. My past research has examined the relationships between image statistics, such as colour, brightness and texture within natural scenes. As useful as this research is, it has to be noted that many factors determine what parts of an image we actually choose to look at. Consequently, we can distinguish between the set of all possible statistics, and the set of image statistics that will fall on human retina. My research will investigate human eye movements to isolate the parts of images that are commonly viewed by humans. These parts will then be analyzed using computational models of the brain to determine how (and if) the relationships between colour, luminance and texture statistics are represented within the visual system. I will investigate how the relationships between these image statistics might be used to signal important image structures, such as edges and shapes that are precursors for more complex tasks, such as object and scene recognition. I will also investigate if our visual system is sensitive to variations in scene statistics using psychophysical experiments, and if such variations have negative consequences for performance in visual tasks such as scene recognition. The findings of my research will advance our understanding of how the human visual perception uses the structures and relationships found in natural images, in order to perform everyday vision tasks.

人类生活的环境是复杂的，但我们的大脑必须在几百毫秒内完成视觉任务，才能生存。我们如何做到这一点是心理学中的一个基本问题。最近的研究表明，我们的环境通过进化和经验塑造了我们感知系统的设计。因此，直接测量自然环境（场景）中的统计量对于促进我们对视觉感知的理解具有巨大的潜在价值。我过去的研究考察了图像统计之间的关系，如自然场景中的颜色，亮度和纹理。尽管这项研究很有用，但必须注意的是，许多因素决定了我们实际上选择查看图像的哪些部分。因此，我们可以区分所有可能的统计数据集和落在人类视网膜上的图像统计数据集。我的研究将调查人类的眼球运动，以分离出人类通常看到的图像部分。然后，这些部分将使用大脑的计算模型进行分析，以确定颜色，亮度和纹理统计数据之间的关系如何（以及是否）在视觉系统中表示。我将研究这些图像统计数据之间的关系如何用于表示重要的图像结构，例如边缘和形状，它们是更复杂任务的前兆，例如对象和场景识别。我还将调查，如果我们的视觉系统是敏感的变化，在现场统计使用心理物理实验，如果这种变化有负面影响的表现在视觉任务，如场景识别。我的研究结果将促进我们对人类视觉感知如何使用自然图像中的结构和关系的理解，以执行日常视觉任务。