基于倒数函数-谱残差的显着对象探测和提取方法A-计算机应用.PDF

Journal of Computer Applications ISSN 1001-9081 算机应用 CODEN JYIIDU 文章编号: 于倒数函数-谱残差的显著对象探测和提取方法 * 陈文兵，鞠虎，陈允杰 (南京信息工程大学 数学与统计学院，南京210044) (*chenwb@) 摘 要: 针对 “中心-周围”的显著对象探测方法频繁出现探测或提取对象不完整，边界不平滑以及其9 级金字塔下采样的 冗余问题，提出一种基于倒数函数-谱残差（Reciprocal Function and Spectral Residual, RFSR ）的显著对象探测方法。该方法首 先利用灰度图像与其对应的高斯低通滤波的差代替 “中心-周围”方法中灰度图像标准化，并减少高斯金字塔至6 级以降低冗 余；其次利用倒数函数滤波器代替Gabor 滤波器提取局部方向信息；接着利用谱残差方法提取图像的谱特征；最后将这三个 特征经过适当融合生成最终显著图。在两个常用基准数据集上的实验结果表明，所提出的倒数-谱残差方法在准确率(precision) 、 召回率 （recall ）及F-measure 等指标上均比 “中心-周围”及谱残差模型有明显提高，其为进一步图像分析、对象识别及基于 显著视觉关注的图像检索等理论及应用研究奠定了基础。 关键词: 显著对象；显著性区域；特征提取；倒数函数；显著图 中图分类号:TP391.41 图像识别及其装置 文献标志码: A A salient object detection and extraction method based on reciprocal function and spectral residual * CHEN Wenbing ，JU Hu ，CHEN Yunj ie (School of Mathematics and Statistics, Nanji ng University of Information Science Technology, Nanji ng210044) Abstract: In order to solve these issues, such as frequently occurring to detect or extract part of obj ect, not smooth boundary and redundant caused by down-sampling 9-level pyramids，a salient obj ect detection method based on reciprocal function and spectral residual (RFSR) was proposed. Firstly, the difference between the intensity image and its corresponding Gaussian low-pass one was used to substitute the normalization of the intensity image under center-surround” model, meanwhile the number of Gaussian pyramids by 6 levels was it further reduced in order to avoid redundant. Secondly, a reciprocal function filter was used to extrac