抗模糊的图像局部特征描述子

doi:10.19665/j.issn1001-2400.2019.01.007

摘要/Abstract

摘要：

针对提取图像局部特征时,尺度不变特征变换描述子对光照条件变化仅有部分不变性,特别对非线性光照变化不具备不变性,对模糊的目标图像也无法准确提取或仅能提取到很少特征点的问题,利用局部二值模式描述子对光照的健壮性提出了一种符合人类视觉系统的自底向上再到自顶向下的视觉认知过程的新的抗模糊的图像局部特征描述子。实验表明,所提出的描述子对光照变化有更好的健壮性,对模糊的目标图像能准确地提取出更多的特征点,保留了尺度不变特征变换对缩放、旋转和压缩等变换的不变性,并显著地提高了针对模糊图像的匹配率。

关键词: 特征描述子, 特征点匹配, 图像匹配, 目标识别

Abstract:

The SIFT descriptor is only partially invariant to illumination when extracting the local features of the image. In particular, the SIFT descriptor is not invariant to non-linear illumination changes and cannot accurately extract the feature points or few of them can be extracted from the fuzzy object image. In order to solve these problems, a new anti-fuzzy local feature descriptor is proposed that is consistent with the visual cognition process of the human visual system from bottom-top and top-down. Experimental results suggest that the proposed operator is robust to the changes of illumination conditions, and more feature points can be extracted accurately from the fuzzy object image. The proposed operator retains the advantages of SIFT descriptors such as invariance of scaling, rotation and compression, and can significantly improve the matching rate on fuzzy images.

Key words: feature descriptor, feature point matching, image matching, object recognition

中图分类号:

TP391.4

唐国良. 抗模糊的图像局部特征描述子[J]. 西安电子科技大学学报, 2019, 46(1): 39-45.

TANG Guoliang. Anti-fuzzy local feature descriptor on images[J]. Journal of Xidian University, 2019, 46(1): 39-45.

图/表 8

图1

图2

表1

图3

表2

图4

表3

图5

参考文献 17

[1]	LOWE D G . Distinctive Image Features from Scale-invariant Keypoints[J]. International Journal of Computer Vision, 2004,60(2):91-110. doi: 10.1023/B:VISI.0000029664.99615.94
[2]	MA J, ZHOU H, ZHAO J , et al. Robust Feature Matching for Remote Sensing Image Registration via Locally Linear Transforming[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015,53(12):6469-6481. doi: 10.1109/TGRS.2015.2441954
[3]	李龙, 刘峥 . 采用多特征联合学习的噪声稳健HRRP识别方法[J]. 西安电子科技大学学报, 2018,45(4):57-62.
	LI Long, LIU Zheng . Noise-robust Multi-feature Joint Learning HRRP Recognition Method[J]. Journal of Xidian University, 2018,45(4):57-62.
[4]	XU X, ZHAO Y. Multimodal Face Recognition for Profile Views Based on SIFT and LBP [C]//Lecture Notes in Computer Science: 8912. Heidelberg: Springer Verlag, 2015: 20-30.
[5]	CHIEN H J, CHUANG C C, CHEN C Y, et al. When to Use What Feature? SIFT, SURF, ORB, or A-KAZE Features for Monocular Visual Odometry [C]//Proceedings of the 2016 International Conference on Image and Vision Computing. Washington: IEEE Computer Society, 2017: 7804434.
[6]	李翠芸, 王精毅, 姬红兵 , 等. 模型参数未知时的CPHD多目标跟踪方法[J]. 西安电子科技大学学报, 2017,44(2):37-41. doi: 10.3969/j.issn.1001-2400.2017.02.007
	LI Cuiyun, WANG Jingyi, JI Hongbing , ed al. CPHD Multi-target Tracking Algorithm with Unknown Model Parameters[J]. Journal of Xidian University, 2017,44(2):37-41. doi: 10.3969/j.issn.1001-2400.2017.02.007
[7]	TAREEN S A K, SALEEM Z. A Comparative Analysis of SIFT, SURF, KAZE, AKAZE, ORB, and BRISK [C]//Proceedings of the 2018 International Conference on Computing, Mathematics and Engineering Technologies: Invent, Innovate and Integrate for Socioeconomic Developmen. Piscataway: IEEE 2018: 1-10.
[8]	LINDEBERG T . Scale Invariant Feature Transform[J]. Scholarpedia, 2012,7(5):10491. doi: 10.4249/scholarpedia.10491
[9]	KABBAI L, AZAZA A, ABDELLAOUI M, et al. Image Matching Based on LBP and SIFT Descriptor [C]//Proceedings of the 2015 12th International Multi-Conference on Systems, Signals and Devices. Piscataway: IEEE, 2015: 7348116.
[10]	BAI S, HOU J, OHNISHI N . Scene Categorization Through Combining LBP and SIFT Features Effectively[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2016,30(1):1655001. doi: 10.1142/S0218001416550016
[11]	MIKSIK O, MIKOLAJCZYK K. Evaluation of Local Detectors and Descriptors for Fast Feature Matching [C]//Proceedings of the 2012 International Conference on Pattern Recognition. Piscataway: IEEE, 2012: 2681-2684.
[12]	CHEON S H, EOM I K, MOON Y H . Fast Descriptor Extraction Method for a SURF-based Interest Point[J]. Electronics Letters, 2016,52(4):274-275. doi: 10.1049/el.2015.3055
[13]	WU J, CUI Z, SHENG V S , et al. A Comparative Study of SIFT and its Variants[J]. Measurement Science Review, 2013,13(3):122-131. doi: 10.2478/msr-2013-0021
[14]	CHENG M M, MITRA N J, HUANG X , et al. Global Contrast Based Salient Region Detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,37(3):569-582. doi: 10.1109/TPAMI.2014.2345401 pmid: 26353262
[15]	GUO Z, ZHANG L, ZHANG D . A Completed Modeling of Local Binary Pattern Operator for Texture Classification[J]. IEEE Transactions on Image Processing, 2010,19(6):1657-1663. doi: 10.1109/TIP.2010.2044957 pmid: 20215079
[16]	SHANNON C E . A Mathematical Theory of Communication[J]. ACM SIGMOBILE Mobile Computing and Communications Review, 2001,5(1):3-55.
[17]	SHANG H, WANG L, HIROSHI T, et al. Character Region Segmentation Based on Stroke Stable Regions [C]//Proceedings of the 2016 International Conference on Pattern Recognition. Piscataway: IEEE, 2016: 3975-3980.

图像尺寸	描述子	(a) bikes	(b) trees	(c) graf	(d) wall	(e) bark	(f) boat	(g) leuven	(h) ubc
128×128	FSIF	5/5	5/5	3/5	3/5	2/5	4/5	5/5	5/5
	SIFT	5/5	5/5	3/5	3/5	1/5	4/5	5/5	5/5
256×256	FSIF	5/5	5/5	4/5	4/5	3/5	4/5	5/5	5/5
	SIFT	5/5	5/5	3/5	4/5	3/5	4/5	5/5	5/5

标准测试数据集	SURF	FAST	SIFT	ORB	MSER	FSIF
DTU ( 64×64 size)	0.04	0	100	0	1.99	100
DTU (128×128 size)	4.03	0.40	4.47	100	1.15	100

模糊图像数据集	SURF	FAST	SIFT	ORB	MSER	FSIF
Lenna (128×128)	9/10	5/10	4/10	10/10	9/10	9/10
Lenna(256×256)	9/10	2/10	5/10	10/10	8/10	8/10
autumn (128×128)	5/10	2/10	4/10	5/10	1/10	9/10
autumn (256×256)	6/10	0/10	6/10	4/10	0/10	10/10