作者:李瑶, 吴国新, 赵西伟, 左云波
作者单位:北京信息科技大学 现代测控技术教育部重点实验室,北京 100192
关键词:SNIC超像素分割;图像特征;显著图;目标检测
摘要:
针对显著性目标检测算法在面对复杂场景图像时存在背景错误凸显和不能凸显纹理细节的问题,提出一种改进的基于SNIC超像素融合纹理特征和上下文内容的显著性目标检测算法。对图像进行SNIC超像素分割,基于超像素的距离、颜色得到超像素显著图;利用LBP算子得到图像的纹理显著图,将超像素显著图以及纹理显著图结合得到初级显著图,最后将初级显著图与CA模型下的显著图融合,得到最终显著图。基于MSRA10K数据集将该文算法与其他5种算法进行对比测试。结果表明:该文所提出的算法与其他现有的显著目标检测算法相比,该文算法可有效抑制相似背景的干扰,突出目标区域的纹理信息,并且具有较高的准确度和稳定性。
Saliency target detection algorithm based on SNIC superpixel fusing texture feature and context content
LI Yao, WU Guoxin, ZHAO Xiwei, ZUO Yunbo
The Ministry of Education Key Laboratory of Modern Measurement and Control Technology, Beijing Information Science & Technology University, Beijing 100192, China
Abstract: Aiming at the problems of background error highlighting and texture details failing to highlight in salient target detection algorithm when facing complex scene images, an improved salient target detection algorithm based on SNIC superpixel fusion of texture features and context content was proposed. Firstly, SNIC superpixel segmentation was performed on the image, and then superpixel salient image was obtained based on the distance and color of the superpixel. The LBP operator is used to get the texture saliency image of the image, and the superpixel saliency image and the texture saliency image are combined to get the primary saliency image. Finally, the primary saliency image is fused with the saliency image under the CA model to get the final saliency image. The proposed algorithm is compared with the other four algorithms based on MSRA10K data set. The results show that compared with other existing salient target detection algorithms, the proposed algorithm can effectively suppress the interference of similar background, highlight the texture information of the target region, and has higher accuracy and stability.
Keywords: SNIC superpixel segmentation;image feature;salience map;target detection
2023, 49(2):93-98 收稿日期: 2021-04-29;收到修改稿日期: 2021-06-22
基金项目: 国家重点研发计划(2020YFB1713200);北京市教委科研计划项目(KM202011232001)
作者简介: 李瑶(1995-),女,山东潍坊市人,硕士研究生,专业方向为图像信息处理
参考文献
[1] 朱斐文, 肖菁, 戴晨, 等. 一种高效的基于局部特征的显著区域检测算法[J]. 计算机应用与软件, 2012, 29(8): 5-9
[2] 刘桂雄, 刘思洋, 吴俊芳, 等. 基于深度学习的机器视觉目标检测算法及在票据检测中应用[J]. 中国测试, 2019, 45(5): 1-9
[3] 陈海永, 郄丽忠, 杨德东, 等. 基于超像素信息反馈的视觉背景提取算法[J]. 光学学报, 2017, 37(7): 186-194
[4] 刘振, 姜晖, 王粒宾. 基于稀疏表示的 SAR 图像目标识别方法[J]. 计算机工程与应用, 2014, 50(10): 212-215
[5] 周莉莉, 姜枫. 图像分割方法综述研究[J]. 计算机应用研究, 2017, 34(7): 1921-1928
[6] 李蔷. 基于内容自适应稀疏字典的图像集压缩算法[J]. 计算机与现代化, 2017(11): 62-66
[7] LI G, YU Y. Visual saliency detection based on multiscale deep CNN features[J]. IEEE Transactions on Image Processing, 2016, 25(11): 5012-5024
[8] 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
[9] BRUCE NDB, TSOTSOS J K. Saliency attention and visual search: an information theoretic approach[J]. Journal of Vision, 2009(3): 1-24
[10] HAREL J, KOCH C, PERONA P. Graph-based visual saliency[C]// Proceedings of the Twentieth Annual Conference on Neutral Information Processing Systems, Vancouver, 2006: 545-552.
[11] HOU X, ZHANG L. Saliency detection: A spectral residual approach[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2007: 1-8.
[12] CHENG M M, ZHANG G X, MITRA N J, et al. Global contrast based salient region detection[J]. Computer Vision & Pattern Recognition, 2011, 37(3): 409-416
[13] STAS G, LIHI ZELN, AYELLET T. Context-aware saliency detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(10): 1915-1926
[14] 丁祖萍, 刘坤, 王成, 等. 一种基于颜色和纹理的显著性目标检测算法[J]. 计算机工程与应用, 2016, 52(19): 192-195
[15] 刘峰, 沈同圣, 韩艳丽, 等. 融合背景感知和颜色对比的显著性检测方法[J]. 计算机辅助设计与图形学学报, 2016, 28(10): 1705-1702