项目名称: 基于视觉显著性结构的特征提取和图像检索
项目编号: No.61202272
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 刘广海
作者单位: 广西师范大学
项目金额: 23万元
中文摘要: 图像检索是人工智能和模式识别领域的一个研究热点。本项目以特征提取和图像检索为研究对象,主要研究内容包括生物视觉特征提取,视觉显著性结构检测,视觉词汇的形成和特征提取。拟采用生物视觉以及模式识别领域最新研究成果,提出一种融合均匀颜色信息的复合感受野模型应用于边缘检测;提出一种融合视觉显著性和局部相似性的结构映射方法;提出视觉结构短语来描述视觉词汇的局部结构特征,并应用于图像检索。拟解决两个关键科学问题:如何结合均匀颜色来模拟复合感受野的抑制和易化作用;如何利用上下文信息来缩小视觉词汇的歧义性。本项目提出的视觉结构短语是视觉词汇的升华,能够为缩小"语义鸿沟"和视觉词汇的歧义性提供新思路。
中文关键词: 图像检索;颜色感知;共生矩阵;方向选择性;视觉注意模型
英文摘要: Image retrieval is a hot topic in the field of artificial intelligence and pattern recognition. The research objects are feature extraction and image retrieval. The main works include biological visual feature extraction, visual salient structure detection, visual words generation and feature extraction. We adopt the recent achievements of biological vision and pattern recognition, and present a new edge detection method fusing uniform color information into the compound receptive field; then put forward the structure mapping method integrating the visual saliency and local similarity; finally, the visual structure phrase is proposed to represent the local structure feature of visual words, and applied to image retrieval. We plan to solve two key scientific problems, one is how to simulate the inhibition and facilitation mechanisms of the compound receptive field involving fusion of uniform color information together, the other is how to reduce the ambiguity of visual words by using contexts information. The proposed visual structure phrase is a significant improvement over visual words. It provides new idea on how to reduce the semantic gap and the ambiguity of visual words.
英文关键词: image retrieval;color perception;co-occurrence matrices;orientation selectivity;visual attention model