论⽂笔记（关于图像检索的总结性论⽂）：Content-
BadImageRetrieva。。。
放上引⽤：Latif, Afshan and Rasheed, Aqsa and Sajid, Umer and Jameel, Ahmed and Ali, Nouman and Ratyal, Naeem Iqbal and Zafar, Bushra and Dar, Saadat and Sajid, Muhammad and Khalil, Tehmina：Content-Bad Image Retrieval and Feature Extraction: A Comprehensive Review，Mathematical Problems in Engineering，Mathematical Problems in Engineering
这是巴基斯坦的⼀个团队的研究论⽂，因为⽆意中看到其实还挺全⾯且详细的。⼀切论⽂都不是完全正确且最新的，这⾥就当和⼤家⼀起基于这篇论⽂重新整理⼀下关于Content-Bad 图像检索和特征抽取的种种。然后也是为了跟着这篇⽂章的参考⽂献思路，各取所需。以下内容仅代表个⼈观点，有问题欢迎交流。
关于什么叫 content-bad，参考以下论⽂：
Gudivada, Venkat N., and Vijay V. Raghavan. "Content-bad image retrieval systems." Computer 28.9 (1995): 18-22.
先看abstract我们可以知道作者写这⼀篇⽂章的⽬的是：
We analyzed the main aspects of various image retrieval and image reprentation models from low-level feature extraction to recent mantic deep-learning approaches. )e important concepts and major rearch studies bad on CBIR and image reprentation are discusd in detail, and future rearch directions are concluded to inspire further rearch in this area.
⽐起以前使⽤metadata以及图像描述的检索，近年CBIR的技术得到了发展，然后他们这篇论⽂呢，就是为了总结从低level的特征抽出图像表现到近年的基于深度学习的图像描述和检索技术，基于图像内容解析的研究，包括对未来这个领域⾛向的⼀些预想。
下⾯进⼊introduction：
作者说，现在很多检索是基于图像描述以及⽤户query的关键词匹配，⽐如以下⼏篇论⽂：
[4] S. Yang, L. Li, S. Wang, W. Zhang, Q. Huang, and Q. Tian,“SkeletonNet: a hybrid network with a skeleton-
embedding process for multi-view image reprentation learning,” IEEETransactions on Multimedia, vol. 1, no. 1, 2019.
[5] W. Zhao, L. Yan, and Y. Zhang, “Geometric-constrained multi-view image matching method ba
d on mi-
global optimization,” Geo-Spatial Information Science, vol. 21, no. 2,pp. 115–126, 2018.
org/abs/1706.06064.
然后作者介绍了CBIR的基础概念和所使⽤的特征，然后作者叙述了特征选择的背景意义：
According to theliterature, the lection of visual features for any system is dependent on the requirements of the end ur.
具体的特征选择还要看⽤户端的需求，为了提⾼检索效果可能很会消耗很⾼的计算成本：
[19] N. Ali, Image Retrieval Using Visual Image Features and Automatic Image Annotation, University of Engineering
大学英语四级考试技巧and
Technology, Taxila, Pakistan, 2016.
[20] B. Zafar, R. Ashraf, N. Ali et al., “Intelligent image classification-bad on spatial weighted histog
rams of
concentric
circles,” Computer Science and Information Systems, vol. 15, no. 3, pp. 615–633, 2018.
不正确的特征选择反⽽会影响系统的表现⽐如：
[12]L. Piras and G. Giacinto, “Information fusion in content bad image retrieval: a comprehensive overview,”
Information平凡的一天作文
Fusion, vol. 37, pp. 50–60, 2017.
然后作者也提了现在各种特征可以⼴泛被运⽤在机器学习和深度学习之中⽽收获好的效果：
ML：
[1] D. Zhang, M. M. Islam, and G. Lu, “A review on automatic image annotation techniques,” Pattern Recognition, vol.
45, no. 1, pp. 346–362, 2012.
[2] Y. Liu, D. Zhang, G. Lu, and W.-Y. Ma, “A survey of contentbad image retrieval with high-level mantics,”
Pattern Recognition, vol. 40, no. 1, pp. 262–282, 2007.
DL（作者也吐槽了句计算消耗⽐较⼤）：
[21] G. Qi, H. Wang, M. Haner, C. Weng, S. Chen, and Z. Zhu,“Convolutional neural network bad detection and
judgement of environmental obstacle in vehicle operation,” CAAI Transactions on Intelligence Technology, vol. 4, no.
2,pp. 80–91, 2019.
[22] U. Markowska-Kaczmar and H. Kwa´snicka, “Deep learning––a new era in bridging the mantic gap,” in
Bridging the Semantic Gap in Image and Video Analysis, pp. 123–159, Springer, Bal, Switzerland, 2018.天山职业技术学院>分手的句子
[23] F. Riaz, S. Jabbar, M. Sajid, M. Ahmad, K. Naer, and N. Ali,“A collision avoidance scheme for autonomous
vehicles inspired by human social norms,” Computers & Electrical Engineering, vol. 69, pp. 690–704, 2018.
所以作者表⽰，这篇⽂章的⼀⼤⽬标就是综合总结分析⼀下各种各样的特征：底层特征（⼏何纹理⾊彩等）会怎样影响检索的效果？如何缩⼩图像底层表现和⾼层语意表现的沟壑？图像的空间布局对图像的检索和表现有多么重要？DL ，ML的导⼊会怎样的提⾼CBIR的表现？
然后作者介绍了下⽂章结构：
=================================================================================
Section 2 颜⾊特征
Section 3 纹理特征
Section 4 形状特征
Section 5 空间特征
Section 6 底层特征融合
买二送一Section 7 局部特征
Section 8 基于深度学习的检索
Section 9 关于⼈脸识别的特征抽出
Section 10 关于距离计算
Section 11 关于特征抽出和CBIR的评价标准
Section 12 关于相关技术的未来
=================================================================================
考虑到阅读的疲惫可能性，本笔记分上中下三部分构成，以上红⾊的内容在（上）部分放置
（以下内容对2015之后的论⽂引⽤会放上链接）
Section 2 关于颜⾊特征：
[24] H. Shao, Y. Wu, W. Cui, and J. Zhang, “Image retrieval bad on MPEG-7 dominant color descriptor,” in
Proceedings of the 9th International Conference for Young Computer Scientists ICYCS 2008, pp. 753–757, IEEE, Hunan, China,November 2008.
基于MPEG-7 descriptor，每个图选8个主⾊，然后基于直⽅图计算图像类似伤心图片带字的
[25] X. Duanmu, “Image retrieval using color moment invariant,”in Proceedings of the 2010 Seventh International
Conference on Information Technology: New Generations (ITNG),pp. 200–203, IEEE, Las Vegas, NV, USA, April 2010.
⽤了HAC聚类颜⾊特征
[26] X.-Y. Wang, B.-B. Zhang, and H.-Y. Yang, “Content-badimage retrieval by integrating color and texture
features,”Multimedia Tools and Applications, vol. 68, no. 3, pp. 545–569, 2014.
对心碰撞⽤了纹理和颜⾊，然后距离计算和合并两种特征造成了难题（以及计算成本）
[27] H. Zhang, Z. Dong, and H. Shu, “Object recognition by acomplete t of pudo-Zernike moment invariants,”
in Proceedings of the 2010 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), pp.
930–933, IEEE, Dallas, TX, USA, March 2010.
基于Zernike 和 pudo-Zernike polynomials拟合的优化来解决缩放旋转问题
作者提到，颜⾊特征是⼀种很难被图像基础形变（旋转，缩放，平移等）所影响的特征，⽐如以下：
[28] J. M. Guo, H. Pratyo, and J. H. Chen, “Content-bad image retrieval using error diffusion block truncation
coding features,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 25, no. 3, pp. 466–481, 2015.
使⽤error diffusion block truncation coding (EDBTC)抽出特征，也就是抽出了颜⾊特征和bitmap特征后进⾏检索
[29] Y. Liu, D. Zhang, and G. Lu, “Region-bad image retrieval with high-level mantics using decision tree
learning,”Pattern Recognition, vol. 41, no. 8, pp. 2554–2570, 2008.
（虽然⽼，这篇想稍微推荐⼀下）这篇论⽂使⽤了决策树，
[30] M. M. Islam, D. Zhang, and G. Lu, “Automatic categorization of image regions using dominant color bad
vector quantization,” in Proceedings of the Digital Image Computing:Techniques and Applications, pp. 191–198, IEEE, Canberra,Australia, December 2008.
这篇是提出了⼀种基于颜⾊的量化⽅法
[31] Z. Jiexian, L. Xiupeng, and F. Yu, “Multiscale distance coherence vector algorithm for content-bad image
retrieval,”@e Scientific World Journal, vol. 2014, Article ID 615973,13 pages, 2014.（虽然个⼈觉得这篇主要是基于轮廓特征，然后经过⼀系列演算实现抗旋转等⼲扰）
然后作者总结，颜⾊特征虽然不能很好的表现局域特征，但是，相对很多区域特征，确实减少了计算消耗然后⽂中给出以上⽅法的检索效率：
在相同datat上来看，【30】提出的颜⾊量化⽅法可以多关注关注。
接下来总结纹理特征：
[32] G. Papakostas, D. Koulouriotis, and V. Tourassis, “Feature extraction bad on wavelet moment
s and moment
invariants in machine vision systems,” in Human-Centric Machine Vision, InTech, London, UK, 2012.
基于⼩波矩和不变矩的特征抽出
[33] G.-H. Liu, Z.-Y. Li, L. Zhang, and Y. Xu, “Image retrieval bad on micro-structure descriptor,” Pattern
Recognition,vol. 44, no. 9, pp. 2123–2133, 2011.
这篇⽂章作者提出了⼀种micro-structures，把HSV⾊彩的特征，和边缘⽅向特征（⽤的Sobel operator）拿来定义了新的特征map
[34] X.-Y. Wang, Z.-F. Chen, and J.-J. Yun, “An effective method for color image retrieval bad on texture,”
Computer Standards & Interfaces, vol. 34, no. 1, pp. 31–35, 2012.
⽤ color co-occurrence matrix 抽出纹理特征
[40] N.-E. Lasmar and Y. Berthoumieu, “Gaussian copula multivariate modeling for texture image retrieval using
wavelet transforms,” IEEE Transactions on Image Processing, vol. 23,no. 5, pp. 2246–2261, 2014.
这篇如标题 wavelet transforms
然后作者总结，因为纹理特征代表的是⼀个像素群，所以它⽐颜⾊特征要更加的具有语意上的意义，但是呢纹理特征有⼀点就是它对噪声很敏感。以上的检索效率如下图：
貔貅挂件
