项目名称: 视觉注意的计算模型及其应用
项目编号: No.61272027
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 王亦洲
作者单位: 北京大学
项目金额: 61万元
中文摘要: 选择性注意是一个非常复杂的感知、认知过程,它时时刻刻影响着大脑的信息加工过程。对选择性注意的研究自上个世纪八十年代以来一直是认知科学的热点领域。我们在本项目中的研究目标是借鉴神经科学、心理物理学和认知科学的研究成果,针对视觉注意的两大任务- - 不带任务的自由扫视和任务驱动的目标搜索,提出一套提出具有创新意义的计算理论、模型和方法,模拟大脑对(视觉)信息的"选择"这一重要功能,通过计眼动实验,在一般自然环境下验证、比较模型的准确性。同时,将研究成果应用于图像和视频的复用(retargeting)、(前景)物体的检测与分割、以及形状分析等实际问题中,以提高它们的效果。
中文关键词: 视觉注意;计算视觉;图像解析;目标检测;
英文摘要: Visual attention is a complex perceptual and cognitive process that selects a limited amount of information out of a huge volume of visual input so that it enables human beings apprehending complex scenes quickly and proficiently. Visual attention has been an active research area since 1980s. In this proposal, we study computational models for visual attention and propose computational theories to account for two of the key visual attentional behaviors, saccades and visual search, from an information theoretic viewpoint based on observed neural science, psychophysics and cognitive science evidences and conclusions. We will simulate and verify eye movements predicted by the proposed theories and models under the two different visual tasks. We will also show that by leveraging the proposed computational theories and models for visual attention, applications, such as image/video retargeting, (foreground/salient) object detection and segmentation, and shape analysis, will be improved.
英文关键词: visual attention;computational vision;image parsing;object detection;