随机颜色直方图的Fisher信息融合及应用.pdfVIP

18 2011 ，47 （1） Computer Engineering and Applications 计算机工程与应用 随机颜色直方图的Fisher 信息融合及应用 1，2 1 1，2 1，2 邹 健 ，刘传才 ，张 玥 ，卢桂馥 1，2 1 1，2 1，2 ZOU Jian ，LIU Chuancai ，ZHANG Yue ，LU Guifu 1.南京理工大学 计算机学院，南京 210094 2.安徽工程大学 应用数理学院，安徽 芜湖 241000 1.School of Computer Science and Technology ，Nanjing University of Science and Technology ，Nanjing 210094 ，China 2.School of Mathematics and Physics ，Anhui Polytechnic University ，Wuhu ，Anhui 241000 ，China E-mail ：qingwenzou@ ZOU Jian ，LIU Chuancai ，ZHANG Yue ，et al.Fisher information fusion of random color histograms and its applications. Computer Engineering and Applications ，2011 ，47 （1）：18-21. Abstract ：The random partition scheme on color space is adopted to extract normalized random color histograms ，thereby re- alizing the extraction of color distribution information of image in dynamic manner.Based on multinomial Fisher geometry ， the averaged extended Fisher information distances are proposed for quantifying overall information divergence between sets of random color histograms.Besides providing the most straightforward elements for k-Nearest Neighbor （kNN ）based image classification ，the measured information divergences also make it possible to learn the collection of sets of random color his- tograms on the closure of a product simplex.Original color-based tasks such as classification ，clustering and visualization on image set can be then implemented on the reduced set.The experimental results on partial images of ALOI database show the effectiveness of proposed method. Key words ：random color histograms ；multinomial geometry ；information fusion ；manifold learning 摘 要：采用在颜色空间的随机划分基础上提取随机直方图从而实现一种以动态方式提取图像颜色分布信息的方法。基于多项 费歇尔几何，采用平均延拓的费歇尔信息距离量化随机颜色直方图集合之间综合的信息差异。除了为基于k-近邻的图像分类提 供最直接判别元素外，融合的平均信息差异也使得在一个乘积多项流形闭包上学习随机颜色直方图集的集合成为可能。基于颜 色特征的图像分类，聚类及可视化等任务此时可在减少集上完成。在ALO