Perceptual Grouping by Position-Frequency Analysis

9319153张本研究以位置-频率分析理论为基础，研究了计算机视觉中的知觉分组方法。感知分组将输入图像的局部特征(如像素、线段和区域)组织成感知显著的结构。对于各种计算机视觉和图像处理应用，例如对象识别、图像分析和主动视觉，成功的感知分组可以导致搜索空间的急剧减少和显著的性能改进。然而，知觉分组的主要困难在于缺乏对如何在适当的数据表示下将知觉分组表示为计算问题的理解。最近时频分析在信号处理中的成功促使了它的扩展--位置频率分析在计算机视觉中的使用。位置-频率表示可用于对广泛的图像特征进行建模，包括局部和全局。它还提供了关于如何将知觉分组表述为计算问题的新见解。本研究项目的目标是：1.)基于位置-频率分析理论，建立了知觉分组的一般数学框架。使用这个框架和开发分组算法，研究感知分组中的一些基本问题，3)在实际(例如，工业)应用中测试这些算法。

9319153 Zhang This research studies an approach to perceptual grouping in computer vision, using as a basis a theory of position-frequency analysis. Perceptual grouping organizes local features of an input image, such as pixels, line-segments, and regions, into perceptually salient structures. Successful perceptual grouping can lead to a drastic reduction of the search space and significant performance improvements for various computer vision and image processing applications, such as object recognition, image analysis, and active vision. The main difficulty in perceptual grouping, however, lies in the lack of understanding of how perceptual grouping should be formulated as a computational problem under a proper data representation. The recent success of time-frequency analysis in signal processing has motivated the use of its extension, position-frequency analysis, in computer vision. Position-frequency representations can be used to model a wide range of image features, local and global. It also provides new insights on how perceptual grouping can be formulated as a computational problem. The objectives of this research project are to: 1.) Develop a general mathematical framework for perceptual grouping based on the theory of position-frequency analysis, 2.) Investigate a number of fundamental problems in perceptual grouping using this framework and developing grouping algorithms, 3.) Test these algorithms in practical (e.g., industrial) applications.