全局对比和背景先验驱动的显著目标检测
|
摘要
针对传统背景先验方法中背景提取不精确并且背景抑制能力弱的问题,提出了全局对比和背景先验驱动的显著目标检测方法。首先将图像分割为一系列感知均匀的超像素,再由全局颜色对比得到基于全局的显著图并计算得到前景种子点;然后将每个边界超像素与前景种子点做对比,筛选差异性较大的边界超像素作为背景种子点并计算得到基于背景的显著图;最后在融合基于全局和背景显著图的基础上,提出一种多兴趣点高斯模型的方法进一步抑制背景并整体高亮显著区域。在公开的MSRA-1000数据测试集上与6种主流方法进行对比实验,结果表明,所提出的显著性目标检测方法对复杂边界信息具有更强的鲁棒性,并能有效抑制背景噪声。
In order to overcome the problems of imprecise in background extraction and weak ability of anti-background existed in traditional approaches of background priors, a salient object detection method using the global contrast and background priors is proposed in this paper. Firstly, the source image is segmented into a series of perceptually uniform superpixels, a global-based saliency map is then calculated by contrasting global color and the foreground seed is collected. Secondly, those superpixels with large difference are selected as the background seeds by comparing each boundary superpixels with the foreground seeds, and the background-based map is then computed. Finally, based on the integration of background-based and global-based saliency map, a multi foci of interest Gaussian model is proposed to reduce background and highlight salient region. Compared with 6 state-of the-art methods on publicly available benchmark datasets(MSRA-1000), the simulation results demonstrate that the salient object detection approach proposed in this paper performs more robustly in dealing with the complex boundary information and suppressing noise comparing with 6 conventional methods.
In order to overcome the problems of imprecise in background extraction and weak ability of anti-background existed in traditional approaches of background priors, a salient object detection method using the global contrast and background priors is proposed in this paper. Firstly, the source image is segmented into a series of perceptually uniform superpixels, a global-based saliency map is then calculated by contrasting global color and the foreground seed is collected. Secondly, those superpixels with large difference are selected as the background seeds by comparing each boundary superpixels with the foreground seeds, and the background-based map is then computed. Finally, based on the integration of background-based and global-based saliency map, a multi foci of interest Gaussian model is proposed to reduce background and highlight salient region. Compared with 6 state-of the-art methods on publicly available benchmark datasets(MSRA-1000), the simulation results demonstrate that the salient object detection approach proposed in this paper performs more robustly in dealing with the complex boundary information and suppressing noise comparing with 6 conventional methods.
引文
[1]Victor L,Pushmeet K,Carsten R,et al.Image segmentation with abounding box prior[C]//IEEE 12th International Conference on Computer Vision,Kyoto,2009:277-284.
[2]Luo P,Tian Y L,Wang X G,et al.Switchable deep network for pedestrian detection[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition,Columbus,2014:899-906.
[3]Ren Z X,Yang N.Region-based saliency detection and its application in object recognition[J].IEEE Transactions on Circuits and Systems for Video Technology,2014,24(5):769-779.
[4]Itti L,Koch C,Niebur E.A model of saliency-based visual attention for rapid scene analysis[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1998,20(11):1254-1259.
[5]郭迎春,袁浩杰,吴鹏.基于Local特征和Regional特征的图像显著性检测[J].自动化报,2013(8):1214-1224.
[6]Cheng M M,Zhang G X,Mitra N J.Global contrast based salient region detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(3):569-582.
[7]Perazzi F,Krahenbuhl P,Pritch Y,et al.Saliency filters:contrast based filtering for salient region detection[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition,Providence,2012:733-740.
[8]Wei Y C,Wen F,Zhu W,et al.Geodesic saliency using background priors[C]//Proceedings of European Conference on Computer Vision,2012:29-42.
[9]蒋寓文,谭乐怡,王守觉.选择性背景优先的显著性检测模型[J].电子与信息学报,2015,37(1):130-136.
[10]Cao X H,Deng H M,Huang B.Background aware saliency detection[J].Systems Engineering and Electronics,2014,36(8):1668-1672.
[11]张巧荣.利用背景先验的显著性检测算法[J].中国图象图形学报,2016,21(2):165-173.
[12]Achanta R,Shaji A,Smith K,et al.SLIC superpixels compared to state-of-the-art superpixel methods[J].IEEE Transactions on Pattern and Machine Intelligence,2012,34(11):2274-2282.
[13]Otsu N.A threshold selection method from gray-level histograms[J].Automatica,1975,9(1):23-27.
[14]高思晗,张雷,李成龙,等.融合低层和高层特征图表示的图像显著性检测算法[J].计算机辅助设计与图形学学报,2016,28(3):420-426.
[15]Wang W,Cai D,Xu X,et al.Visual saliency detection based on region descriotors and prior knowledeg[J].Signal Processing Image Communication,2014,29(3):424-433.
[16]Achanta R,Hemami S,Estrada F,et al.Frequency-tuned salient region detection[C]//IEEE International Conference on Computer Vision and Pattern Recognition,Miami,2009:1597-1604.
[17]Yang C,Zhang L H,Lu H C.Graph-regularized saliency detection with convex-hull-based center prior[J].IEEE Signal Processing Letters,2013,20(7):637-640.
[18]Ran M,Ayellet T,Lihi Z.What makes a patch distinct[C]//IEEE Conference on Computer Vision and Pattern Recognition,2013:1139-1146.
[19]李岳云,许悦雷,马时平,等.深度卷积神经网络的显著性检测[J].中国图象图形学报,2016,21(1):53-59.
[2]Luo P,Tian Y L,Wang X G,et al.Switchable deep network for pedestrian detection[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition,Columbus,2014:899-906.
[3]Ren Z X,Yang N.Region-based saliency detection and its application in object recognition[J].IEEE Transactions on Circuits and Systems for Video Technology,2014,24(5):769-779.
[4]Itti L,Koch C,Niebur E.A model of saliency-based visual attention for rapid scene analysis[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1998,20(11):1254-1259.
[5]郭迎春,袁浩杰,吴鹏.基于Local特征和Regional特征的图像显著性检测[J].自动化报,2013(8):1214-1224.
[6]Cheng M M,Zhang G X,Mitra N J.Global contrast based salient region detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(3):569-582.
[7]Perazzi F,Krahenbuhl P,Pritch Y,et al.Saliency filters:contrast based filtering for salient region detection[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition,Providence,2012:733-740.
[8]Wei Y C,Wen F,Zhu W,et al.Geodesic saliency using background priors[C]//Proceedings of European Conference on Computer Vision,2012:29-42.
[9]蒋寓文,谭乐怡,王守觉.选择性背景优先的显著性检测模型[J].电子与信息学报,2015,37(1):130-136.
[10]Cao X H,Deng H M,Huang B.Background aware saliency detection[J].Systems Engineering and Electronics,2014,36(8):1668-1672.
[11]张巧荣.利用背景先验的显著性检测算法[J].中国图象图形学报,2016,21(2):165-173.
[12]Achanta R,Shaji A,Smith K,et al.SLIC superpixels compared to state-of-the-art superpixel methods[J].IEEE Transactions on Pattern and Machine Intelligence,2012,34(11):2274-2282.
[13]Otsu N.A threshold selection method from gray-level histograms[J].Automatica,1975,9(1):23-27.
[14]高思晗,张雷,李成龙,等.融合低层和高层特征图表示的图像显著性检测算法[J].计算机辅助设计与图形学学报,2016,28(3):420-426.
[15]Wang W,Cai D,Xu X,et al.Visual saliency detection based on region descriotors and prior knowledeg[J].Signal Processing Image Communication,2014,29(3):424-433.
[16]Achanta R,Hemami S,Estrada F,et al.Frequency-tuned salient region detection[C]//IEEE International Conference on Computer Vision and Pattern Recognition,Miami,2009:1597-1604.
[17]Yang C,Zhang L H,Lu H C.Graph-regularized saliency detection with convex-hull-based center prior[J].IEEE Signal Processing Letters,2013,20(7):637-640.
[18]Ran M,Ayellet T,Lihi Z.What makes a patch distinct[C]//IEEE Conference on Computer Vision and Pattern Recognition,2013:1139-1146.
[19]李岳云,许悦雷,马时平,等.深度卷积神经网络的显著性检测[J].中国图象图形学报,2016,21(1):53-59.