Figure - Ground Segmentation and Shape Description

图-地面分割和形状描述

• 批准号: 9616555
• 负责人: Gordon Baylis
• 金额: $ 21.85万
• 依托单位: University of South Carolina at Columbia
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-09-01 至 2001-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9616555&HistoricalAwards=false
• 关键词: Figure; Ground; Segmentation; Shape; Description

9616555 BAYLIS Though we all take for granted our amazing ability to see, the process of vision is a very complex one. In order to understand vision it is necessary to subdivide it into a number of stages. This project will investigate the first of these stages, in which visual information is sorted or parsed initially. The result of visual parsing is that as we look at the world, we immediately and without effort get an impression of a world that is comprised of things within a scene. This research derives from a theory of visual parsing, due to Baylis and Driver, that says that deciding which parts of a visual scene are objects and determining the shapes of those objects are intermeshed processes. The experiments will test this theory. Imagine that one is looking at a cup on a table. Recognizing that it is a cup requires knowledge that there is a "cup-shaped" object lying on the table. This in turn requires that we appreciate that the outline of the cup "belongs" to the cup; otherwise we might think there is a strange "cup-shaped" hole in the table! Baylis and Driver's theory holds that deciding that the cup-shaped outline "belongs" to the cup, not the table, is a largely automatic process; from this premise, the theory makes specific predictions about which perceptual acts will be easy and which will be hard. The experiments will ask people to make judgments about what they see and measure the speed and accuracy of their judgments; it will thus test the theory of visual parsing. The fact that visual parsing in people is very fast, efficient, and effortless belies the importance of this system. It is not possible to understand the processes of vision and visual recognition without a clear understanding of the initial processes of visual parsing. There are many potential applications of a better understanding of vision, including the creation of automated vision systems for use in industry and medicine. Many attempts at machine vision have faile d primarily because of inadequate visual parsing. For example, an automated system for detecting cancerous cells in a blood sample needs to be able to "find" which regions of a slide are the cells and which regions are the spaces between. Similarly, it is easy to see how catastrophic even a partial breakdown of visual parsing would be to a person's ability to make sense of the visual world. Many neuropsychological disorders in patients with brain damage caused by strokes include impaired visual parsing. Thus, a better understanding of visual parsing and how it relates to other processes in vision will not only lead to a clearer theoretical conception of how it takes place, but can also ultimately lead to a better understanding of how visual parsing breaks down in patients with brain damage, improve the rehabilitation of patients with these problems, and help in the design of machine vision systems. ***

9616555 BAYLIS虽然我们都认为我们拥有惊人的视觉能力，但视觉的过程是一个非常复杂的过程。 为了理解视觉，有必要将其细分为若干阶段。 这个项目将研究这些阶段中的第一个阶段，在这个阶段中，视觉信息最初被排序或解析。 视觉解析的结果是，当我们看世界时，我们立即毫不费力地获得了一个由场景中的事物组成的世界的印象。 这项研究源于Baylis和Driver的视觉解析理论，即决定视觉场景的哪些部分是对象以及确定这些对象的形状是相互交织的过程。 实验将检验这一理论。 想象一个人正在看桌子上的一个杯子。 认识到它是一个杯子需要知道桌子上有一个“杯形”物体。 这反过来又要求我们欣赏杯子的轮廓“属于”杯子;否则我们可能会认为桌子上有一个奇怪的“杯形”洞！ 贝利斯和德赖弗的理论认为，决定杯形轮廓“属于”杯子，而不是桌子，在很大程度上是一个自动的过程;从这个前提出发，该理论对哪些知觉行为容易，哪些知觉行为困难做出了具体的预测。 这些实验将要求人们对他们所看到的东西做出判断，并测量他们判断的速度和准确性;因此，它将测试视觉解析理论。 人类的视觉解析非常快速、高效和毫不费力，这一事实掩盖了这个系统的重要性。 如果不清楚视觉解析的初始过程，就不可能理解视觉和视觉识别的过程。 更好地理解视觉有许多潜在的应用，包括创建用于工业和医学的自动视觉系统。 许多机器视觉的尝试都失败了，主要是因为视觉解析不足。 例如，用于检测血液样本中的癌细胞的自动化系统需要能够“发现”载玻片的哪些区域是细胞，哪些区域是之间的空间。 同样，很容易看出，即使是视觉解析的部分崩溃，对一个人理解视觉世界的能力也会造成多么灾难性的影响。 中风导致脑损伤的患者的许多神经心理障碍包括视觉解析受损。 因此，更好地理解视觉解析以及它与视觉中其他过程的关系不仅会导致对它如何发生的更清晰的理论概念，而且最终还可以更好地理解视觉解析如何在脑损伤患者中分解，改善这些问题患者的康复，并帮助设计机器视觉系统。 ***