【作者】 刘涛；

【导师】 张艳宁；

【作者基本信息】 西北工业大学 ， 计算机应用技术， 2006， 硕士

【摘要】 基于内容的图像检索CBIR(Content-based Image Retrieval)是多媒体信息处理的研究热点之一,有着广泛的应用背景。本文主要针对现有基于内容检索方法对特定的应用有较强的依赖性,特定应用背景中采用不同的特征提取方法对检索效果有较大影响的问题,分别从颜色特征、纹理特征和形状特征三方面详细分析了现有的特征提取方法,并提出了一种改进的形状特征提取方法。另外本文对被广泛关注的基于语义图像检索方法和基于压缩域的图像检索方法进行了相应的研究。在此基础上,构建了一个实用的图像检索系统原型。本文的主要工作有: 在图像特征提取方面,分别从颜色特征、纹理特征和形状特征三个方面进行了分析和研究,并实现了颜色直方图、分块颜色直方图、累计颜色直方图、颜色矩、共生矩阵、Gabor滤波器组、高斯马尔科夫随机场、形状参数、Hu矩和Zernike矩等常见的特征提取方法。在形状特征的提取上,本文提出了一种“基于遗传算法的自适应边界点选取形状上下文”的特征提取方法。实验表明,与传统的形状上下文方法相比该方法具有较好的检索性能。 在语义检索方面,在分析常用的语义检索方法的基础上,本文提出了一种基于区域的图像检索方法,实验证明该方法具有较好的检索效果。另外在相关反馈方面,本文实现了一种加权查询向量修改反馈方法,通过实验证明了该方法的有效性。 在基于压缩域的图像检索方面,对现有的基于压缩域的图像检索技术进行了分析,并介绍了基于JPEG和JPEG2000压缩标准的检索方法,最后实现了一种基于JPEG图像的压缩域纹理图像检索方法,实验证明该方法具有较好的检索性能。 基于以上研究,设计并实现了一个高效、灵活的图像检索原型系统。该原型系统可以方便的利用本文中介绍的各种检索算法进行图像检索,并可以根据用户对检索结果主观评价的程度进行相关反馈。另外,在用户接口上,系统提供直接选择、随机选择和浏览选择三种查询图像选择方法,使用户可以以更加人性化的方式选择查询图像。实验证明,本文实现的系统具有较好的检索性能。更多还原

【Abstract】 Content-based image retrieval (CBIR) is a hotspot of research in multimedia information processing field, and has capacious developed foreground. CBIR seems to be highly task-dependent. In practice, special algorithms are taken according to special application. To slove this problem, this paper analyzed the method of feature extraction, and presented a new method of shape feature extraction. In addition, we study the technology of semantic retrieval and compressed domain image retrieval. On the basis of it, a system prototype of CBIR is designed.On the aspect of feature extraction, we analyzed separately color feature extraction, texture feature extraction and shape feature extraction and realized frequently used methods, like color histogram, co-occurrence matrix and so on. On the basis of shape context, a new shape feature extraction method called Shape context by choosing the edge points self-adaptively based on genetic algorithm is proposed. Experiments show that the method acquired better effect compared with shape context.On the aspect of semantic retrieval, we proposed a new method based on object region on the basis of analyzing frequently used semantic retrieval methods. In addition, to overcome the drawback of query vector modified feedback method, a new method named weighted query vector modified is realized. Experiments show the methods are satisfactory.On the aspect of compressed domain image retrieval, we analyzed the retrieval methods based on JPEG and JPEG2000 on the basis of analyzing frequently used compressed domain image retrieval methods. At last, we realized a method based on JPEG image, Experments show the method aquired better retrieval effect.A system prototype of content-based image retrieval is designed and realized on the basis of it. The system can retrieval image using the methods of this paper, and user can feedback by his subjective evaluation. In the user interface, we designed three query image select method. By this way, user can select query image easily. Experiments show that the system acquired better retrieval effect.更多还原