基于图割精细化和可微分聚类的无监督显著性目标检测

doi:10.11772/j.issn.1001-9081.2021061054

《计算机应用》唯一官方网站 ›› 2021, Vol. 41 ›› Issue (12): 3571-3577.DOI: 10.11772/j.issn.1001-9081.2021061054

所属专题：第十八届中国机器学习会议(CCML 2021)

• 第十八届中国机器学习会议(CCML 2021) • 上一篇下一篇

基于图割精细化和可微分聚类的无监督显著性目标检测

李小雨¹^,²^,³, 房体育¹^,²^,³, 夏英杰¹^,²^,³, 李金屏¹^,²^,³()

^1.济南大学信息科学与工程学院，济南 250022
^2.山东省网络环境智能计算技术重点实验室（济南大学），济南 250022
^3.山东省“十三五”高校信息处理与认知计算重点实验室（济南大学），济南 250022

收稿日期:2021-05-12 修回日期:2021-07-27 接受日期:2021-08-05 发布日期:2021-12-28 出版日期:2021-12-10
通讯作者: 李金屏
作者简介:李小雨（1996—），女，四川内江人，硕士研究生，CCF会员，主要研究方向：图像处理、机器学习
房体育（1995—），男，山东济宁人，硕士研究生，CCF会员，主要研究方向：图像处理、机器学习
夏英杰（1978—），男，山东威海人，博士，讲师，CCF会员，主要研究方向：图像处理、机器学习；
基金资助:
山东省重点研发计划项目(2017CXGC0810);山东省高等学校科技发展计划项目(J18KA371)

Unsupervised salient object detection based on graph cut refinement and differentiable clustering

Xiaoyu LI¹^,²^,³, Tiyu FANG¹^,²^,³, Yingjie XIA¹^,²^,³, Jinping LI¹^,²^,³()

^1.School of Information Science and Engineering，University of Jinan，Jinan Shandong 250022，China
^2.Shandong Provincial Key Laboratory of Network Based Intelligent Computing （University of Jinan），Jinan Shandong 250022，China
^3.Shandong College and University Key Laboratory of Information Processing and Cognitive Computing in the 13th Five-Year Plan （University of Jinan），Jinan Shandong 250022，China

Received:2021-05-12 Revised:2021-07-27 Accepted:2021-08-05 Online:2021-12-28 Published:2021-12-10
Contact: Jinping LI
About author:LI Xiaoyu， born in 1996， M. S. candidate. Her research interests include image processing， machine learning.
FANG Tiyu， born in 1995， M. S. candidate. His research interests include image processing， machine learning.
XIA Yingjie， born in 1978， Ph. D.， lecturer. His research interests include image processing， machine learning.
Supported by:
the Shandong Provincial Key Research and Development Program(2017CXGC0810);the Shandong Province Higher Educational Science and Technology Development Program(J18KA371)

摘要/Abstract

摘要：

针对传统显著性检测算法分割精度低以及基于深度学习的显著性检测算法对像素级人工注释数据依赖性过强等不足，提出一种基于图割精细化和可微分聚类的无监督显著性目标检测算法。该算法采用由“粗”到“精”的思想，仅利用单张图像的特征便可以实现精确的显著性目标检测。首先利用Frequency-tuned算法根据图像自身的颜色和亮度得到显著粗图，然后根据图像的统计特性进行二值化并结合中心优先假设得到显著目标的候选区域，进而利用基于单图像进行图割的GrabCut算法对显著目标进行精细化分割，最后为克服背景与目标极为相似时检测不精确的困难，引入具有良好边界分割效果的无监督可微分聚类算法对单张显著图做进一步的优化。所提出的算法在ECSSD和SOD数据集上进行测试并与现有的7种算法进行对比，结果表明得到的优化显著图更接近于真值图，在ECSSD和SOD数据集上分别实现了14.3%和23.4%的平均绝对误差（MAE）。

关键词: 显著性目标检测, Frequency-tuned算法, GrabCut算法, 可微分聚类, 由“粗”到“精”

Abstract:

Concerning that the traditional saliency detection algorithm has low segmentation accuracy and the deep learning-based saliency detection algorithm has strong dependence on pixel-level manual annotation data， an unsupervised saliency object detection algorithm based on graph cut refinement and differentiable clustering was proposed. In the algorithm， the idea of “coarse” to “fine” was adopted to achieve accurate salient object detection by only using the characteristics of a single image. Firstly， Frequency-tuned algorithm was used to obtain the salient coarse image according to the color and brightness of the image itself. Then， the candidate regions of the salient object were obtained by binarization according to the image’s statistical characteristics and combination of the central priority hypothesis. After that， GrabCut algorithm based on single image for graph cut was used for segmenting the salient object finely. Finally， in order to overcome the difficulty of imprecise detection when the background was very similar to the object， the unsupervised differentiable clustering algorithm with good boundary segmentation effect was introduced to further optimize the saliency map. Experimental results show that compared with the existing seven algorithms， the optimized saliency map obtained by the proposed algorithm is closer to the ground truth， achieving an Mean Absolute Error （MAE） of 14.3% and 23.4% on ECSSD and SOD datasets， respectively.

Key words: salient object detection, Frequency-tuned algorithm, GrabCut algorithm, differentiable clustering, from "coarse" to "fine"

中图分类号:

TP391.41

李小雨, 房体育, 夏英杰, 李金屏. 基于图割精细化和可微分聚类的无监督显著性目标检测[J]. 计算机应用, 2021, 41(12): 3571-3577.

Xiaoyu LI, Tiyu FANG, Yingjie XIA, Jinping LI. Unsupervised salient object detection based on graph cut refinement and differentiable clustering[J]. Journal of Computer Applications, 2021, 41(12): 3571-3577.

图/表 11

图1 本文算法流程

Fig. 1 Flowchart of the proposed algorithm

图2 Frequency-tuned算法显著性检测结果

Fig. 2 Saliency detection results of Frequency-tuned algorithm

图3 候选区域选择结果

Fig. 3 Results of candidate region selection

图4 精细化分割图示

Fig. 4 Refined segmentation diagram

图5 可微分聚类算法流程

Fig. 5 Flowchart of differentiable clustering algorithm

图6 可微分聚类优化效果

Fig. 6 Differentiable clustering optimization effect

表1 不同显著性检测算法在ECSSD和SOD数据集上的指标得分

Tab. 1 Index scores of different saliency detection methods on ECSSD and SOD datasets

算法	ECSSD数据集		SOD数据集
算法	$e M A E$	F‑measure	$e M A E$	F‑measure
GBVS^［16］	0.263	0.599	0.292	0.572
RPCA^［17］	0.247	0.573	0.294	0.509
SF^［18］	0.329	0.307	0.329	0.296
SLR^［19］	0.252	0.442	0.248	0.395
RBD^［20］	0.171	0.513	0.229	0.428
FBS^［21］	0.340	0.444	0.381	0.403
SC^［22］	0.190	0.751	0.264	0.643
本文算法	0.143	0.722	0.234	0.572

表1 不同显著性检测算法在ECSSD和SOD数据集上的指标得分

Tab. 1 Index scores of different saliency detection methods on ECSSD and SOD datasets

算法	ECSSD数据集		SOD数据集
算法	$e M A E$	F‑measure	$e M A E$	F‑measure
GBVS^［16］	0.263	0.599	0.292	0.572
RPCA^［17］	0.247	0.573	0.294	0.509
SF^［18］	0.329	0.307	0.329	0.296
SLR^［19］	0.252	0.442	0.248	0.395
RBD^［20］	0.171	0.513	0.229	0.428
FBS^［21］	0.340	0.444	0.381	0.403
SC^［22］	0.190	0.751	0.264	0.643
本文算法	0.143	0.722	0.234	0.572

表2 显著粗图、精细化分割图、优化显著图的比较

Tab. 2 Comparison of saliency rough map， refined segmentation map and optimized saliency map

生成结果	ECSSD数据集				SOD数据集
生成结果	$e M A E$	P	R	F‑measure	$e M A E$	P	R	F‑measure
显著粗图	0.318	0.229	0.278	0.238	0.345	0.248	0.245	0.247
精细化分割图	0.150	0.628	0.763	0.655	0.241	0.497	0.654	0.526
优化显著图	0.143	0.715	0.746	0.722	0.234	0.554	0.640	0.572

表2 显著粗图、精细化分割图、优化显著图的比较

Tab. 2 Comparison of saliency rough map， refined segmentation map and optimized saliency map

生成结果	ECSSD数据集				SOD数据集
生成结果	$e M A E$	P	R	F‑measure	$e M A E$	P	R	F‑measure
显著粗图	0.318	0.229	0.278	0.238	0.345	0.248	0.245	0.247
精细化分割图	0.150	0.628	0.763	0.655	0.241	0.497	0.654	0.526
优化显著图	0.143	0.715	0.746	0.722	0.234	0.554	0.640	0.572

表3 二值化阈值的灵敏度分析（b=10）

Tab. 3 Sensitivity analysis of binarization threshold (b=10)

参数 $a$	ECSSD数据集		SOD数据集
参数 $a$	$e M A E$	F‑measure	$e M A E$	F‑measure
0.40	0.141	0.719	0.239	0.570
0.45	0.143	0.722	0.234	0.572
0.50	0.148	0.714	0.239	0.565
0.55	0.155	0.703	0.248	0.549

表3 二值化阈值的灵敏度分析（b=10）

Tab. 3 Sensitivity analysis of binarization threshold (b=10)

参数 $a$	ECSSD数据集		SOD数据集
参数 $a$	$e M A E$	F‑measure	$e M A E$	F‑measure
0.40	0.141	0.719	0.239	0.570
0.45	0.143	0.722	0.234	0.572
0.50	0.148	0.714	0.239	0.565
0.55	0.155	0.703	0.248	0.549

表4 候选区域选择的灵敏度分析（a=0.45）

Tab. 4 Sensitivity analysis of candidate area selection (a=0.45)

参数 $b$	ECSSD数据集		SOD数据集
参数 $b$	$e M A E$	F‑measure	$e M A E$	F‑measure
5	0.156	0.701	0.247	0.558
10	0.143	0.722	0.234	0.572
15	0.139	0.719	0.230	0.566

表4 候选区域选择的灵敏度分析（a=0.45）

Tab. 4 Sensitivity analysis of candidate area selection (a=0.45)

参数 $b$	ECSSD数据集		SOD数据集
参数 $b$	$e M A E$	F‑measure	$e M A E$	F‑measure
5	0.156	0.701	0.247	0.558
10	0.143	0.722	0.234	0.572
15	0.139	0.719	0.230	0.566

图7 本文优化算法在ECSSD、SOD数据集上的视觉显著图对比

Fig. 7 Comparison of visual saliency maps on ECSSD and SOD datasets

参考文献 22

1	ACHANTA R， HEMAMI S， ESTRADA F， et al. Frequency-tuned salient region detection［C］// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2009：1597-1604. 10.1109/cvpr.2009.5206596
2	VALENTI R， SEBE N， GEVERS T. Image saliency by isocentric curvedness and color［C］// Proceedings of the IEEE 12th International Conference on Computer Vision. Piscataway： IEEE， 2009： 2185-2192. 10.1109/iccv.2009.5459240
3	WANG M， KONRAD J， ISHWAR P， et al. Image saliency： from intrinsic to extrinsic context［C］// Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2011： 417-424. 10.1109/cvpr.2011.5995743
4	CHENG M M， MITRA N J， HUANG X L， et al. Global contrast based salient region detection［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2015， 37（3）： 569-582. 10.1109/tpami.2014.2345401
5	CHEN T S， LIN L， LIU L B， et al. DISC： deep image saliency computing via progressive representation learning［J］. IEEE Transaction on Neural Networks and Learning Systems， 2016， 27（6）： 1135-1149. 10.1109/tnnls.2015.2506664
6	LEE G， TAI Y W， KIM J. Deep saliency with encoded low level distance map and high level features［C］// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2016： 660-668. 10.1109/cvpr.2016.78
7	HOU Q B， CHENG M M， HU X W， et al. Deeply supervised salient object detection with short connections［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2019， 41（4）： 815-828. 10.1109/tpami.2018.2815688
8	WANG L J， LU H C， WANG Y F， et al. Learning to detect salient objects with image-level supervision［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 3796-3805. 10.1109/cvpr.2017.404
9	ZHANG D W， HAN J W， ZHANG Y. Supervision by fusion： towards unsupervised learning of deep salient object detector［C］// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway： IEEE， 2017： 4068-4076. 10.1109/iccv.2017.436
10	ZHANG J， ZHANG T， DAI Y C， et al. Deep unsupervised saliency detection： a multiple noisy labeling perspective［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 9029-9038. 10.1109/cvpr.2018.00941
11	BORJI A， CHENG M M， HOU Q B， et al. Salient object detection： a survey［J］. Computational Visual Media， 2019， 5（2）： 117-150. 10.1007/s41095-019-0149-9
12	ROTHER C， KOLMOGOROV V， BLAKE A. “GrabCut” — interactive foreground extraction using iterated graph cut［J］. ACM Transactions on Graphics， 2004， 23（3）： 309-314. 10.1145/1015706.1015720
13	KIM W， KANEZAKI A， TANAKA M. Unsupervised learning of image segmentation based on differentiable feature clustering［J］. IEEE Transactions on Image Processing， 2020， 29： 8055-8068. 10.1109/tip.2020.3011269
14	YAN Q， XU L， SHI J P， et al. Hierarchical saliency detection［C］// Proceedings of the 2013 IEEE International Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2013： 1155-1162. 10.1109/cvpr.2013.153
15	MOVAHEDI V， ELDER J H. Design and perceptual validation of performance measures for salient object segmentation［C］// Proceedings of the 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2010： 49-56. 10.1109/cvprw.2010.5543739
16	HAREL J， KOCH C， PERONA P. Graph-based visual saliency［C］// Proceedings of the 19th International Conference on Neural Information Processing Systems. Cambridge： MIT Press， 2006： 545-552. 10.7551/mitpress/7503.003.0073
17	CANDÈS E J， LI X D， MA Y， et al. Robust principal component analysis？［J］. Journal of the ACM， 2011， 58（3）： No.11. 10.1145/1970392.1970395
18	PERAZZI F， KRÄHENBÜHL P， PRITCH Y， et al. Saliency filters： contrast based filtering for salient region detection［C］// Proceedings of the 2012 IEEE International Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2012： 733-740. 10.1109/cvpr.2012.6247743
19	ZOU W B， KPALMA K， LIU Z， et al. Segmentation driven low-rank matrix recovery for saliency detection［C］// Proceedings of the 2013 British Machine Vision Conference. Durham： BMVA Press， 2013： No.78. 10.5244/c.27.78
20	ZHU W J， LIANG S， WEI Y C， et al. Saliency optimization from robust background detection［C］// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2014： 2814-2821. 10.1109/cvpr.2014.360
21	PAPALE P， LEO A， CECCHETTI L， et al. Foreground-background segmentation revealed during natural image viewing［J］. eNeuro， 2018， 5（3）： No.ENEURO.0075-18.2018. 10.1523/eneuro.0075-18.2018
22	WU X Y， MA X D， ZHANG J X， et al. Salient object detection via deformed smoothness constraint［C］// Proceedings of the 25th IEEE International Conference on Image Processing. Piscataway： IEEE， 2018：2815-2819. 10.1109/icip.2018.8451169

[1]	徐松, 张文博, 王一帆. 基于时空信息的轻量视频显著性目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2192-2199.
[2]	王强, 黄小明, 佟强, 刘秀磊. 基于边界框标注的弱监督显著性目标检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(6): 1910-1918.
[3]	汪虹余, 张彧, 杨恒, 穆楠. 基于蚁群优化算法的弱光图像显著性目标检测[J]. 计算机应用, 2021, 41(10): 2970-2978.
[4]	周良芬何建农. 基于GrabCut改进的图像分割算法[J]. 计算机应用, 2013, 33(01): 49-52.

基于图割精细化和可微分聚类的无监督显著性目标检测

Unsupervised salient object detection based on graph cut refinement and differentiable clustering

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 22

相关文章 4

编辑推荐

Metrics