基于非局部梯度的圖像質(zhì)量評(píng)價(jià)算法

高敏娟; 黨宏社; 魏立力; 張選德

doi:10.11999/JEIT180597

基于非局部梯度的圖像質(zhì)量評(píng)價(jià)算法

doi: 10.11999/JEIT180597

1.
陜西科技大學(xué)電氣與信息工程學(xué)院 ??西安 ??710021
2.
寧夏大學(xué)數(shù)學(xué)統(tǒng)計(jì)學(xué)院 ??銀川 ??750021

基金項(xiàng)目: 國(guó)家自然科學(xué)基金(61871260, 61603234, 61362029, 61461043)

詳細(xì)信息

作者簡(jiǎn)介:
高敏娟：女，1984年生，博士生，研究方向?yàn)閳D像處理、圖像質(zhì)量評(píng)價(jià)

黨宏社：男，1962年生，教授，博士生導(dǎo)師，研究方向?yàn)楣I(yè)過(guò)程與優(yōu)化、計(jì)算機(jī)控制、圖像處理

魏立力：男，1965年生，教授，研究方向?yàn)閼?yīng)用統(tǒng)計(jì)與數(shù)據(jù)分析

張選德：男，1979 年生，教授，博士生導(dǎo)師，研究方向?yàn)閳D像恢復(fù)、圖像質(zhì)量評(píng)價(jià)、稀疏表示和低秩逼近理論

通訊作者:
張選德　zhangxuande@sust.edu.cn

中圖分類(lèi)號(hào): TP391
計(jì)量
- 文章訪(fǎng)問(wèn)數(shù): 2130
- HTML全文瀏覽量: 973
- PDF下載量: 111
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2018-06-19
- 修回日期: 2018-12-18
- 網(wǎng)絡(luò)出版日期: 2018-12-26
- 刊出日期: 2019-05-01

Image Quality Assessment Algorithm Based on Non-local Gradient

1.
College of Electrical and Information Engineering, Shaanxi University of Science & Technology, Xi’an 710021, China
2.
School of Mathematics and Statistics, Ningxia University, Yinchuan 750021, China

Funds: The National Natural Science Foundation of China (61871260, 61603234, 61362029, 61461043)

摘要

摘要:
圖像質(zhì)量評(píng)價(jià)研究的目標(biāo)在于模擬人類(lèi)視覺(jué)系統(tǒng)對(duì)圖像質(zhì)量的感知過(guò)程，構(gòu)建與主觀評(píng)價(jià)結(jié)果盡可能一致的客觀評(píng)價(jià)算法?，F(xiàn)有的很多算法都是基于局部結(jié)構(gòu)相似設(shè)計(jì)的，但人對(duì)圖像的主觀感知是高級(jí)的、語(yǔ)義的過(guò)程，而語(yǔ)義信息本質(zhì)上是非局部的，因此圖像質(zhì)量評(píng)價(jià)應(yīng)該考慮圖像的非局部信息。該文突破了經(jīng)典的基于局部信息的算法框架，提出一種基于非局部信息的框架，并在此框架內(nèi)構(gòu)建了一種基于非局部梯度的圖像質(zhì)量評(píng)價(jià)算法，該算法通過(guò)度量參考圖像與失真圖像的非局部梯度之間的相似性來(lái)預(yù)測(cè)圖像質(zhì)量。在公開(kāi)測(cè)試數(shù)據(jù)庫(kù)TID2008, LIVE, CSIQ上的數(shù)值實(shí)驗(yàn)結(jié)果表明，該算法能獲得較好的評(píng)價(jià)效果。
- 圖像質(zhì)量評(píng)價(jià) /
- 人類(lèi)視覺(jué)系統(tǒng) /
- 非局部梯度
Abstract:
The goal of Image Quality Assessment (IQA) research is to simulate the Human Visual System’s (HVS) perception process of assessing image quality and construct an objective evaluation algorithm that is as consistent as the subjective evaluation result. Many existing algorithms are designed based on local structural similarity, but human subjective perception of images is a high-level, semantic process, and semantic information is essentially non-local, so image quality assessment should take the non-local information of the image into consideration. This paper breaks through the classical framework based on local information, and proposes a framework based on non-local information. Under the proposed framework, an image quality assessment algorithm based on non-local gradient is also presented. This algorithm predicts image quality by measuring the similarity between the non-local gradients of reference image and the distorted image. The experimental results on the public test database TID2008, LIVE, and CSIQ show that the proposed algorithm can obtain better evaluation results.
- Image Quality Assessment (IQA) /
- Human Visual System (HVS) /
- Non-local gradient

HTML全文

圖 1 基于局部和非局部信息的FRIQA模型兩步框架

下載: 全尺寸圖片幻燈片

圖 2 參考圖像中以$i$為中心、$t$為邊長(zhǎng)的方鄰域

下載: 全尺寸圖片幻燈片

圖 3 6種算法在TID2008數(shù)據(jù)庫(kù)中的散點(diǎn)圖

下載: 全尺寸圖片幻燈片

表 1 10種不同IQA算法在TID2008, CSIQ, LIVE數(shù)據(jù)庫(kù)的實(shí)驗(yàn)結(jié)果比較

數(shù)據(jù)庫(kù)	性能指標(biāo)	PSNR	VSNR	SSIM	MS-SSIM	IW-SSIM	FSIM	ESSIM	GMSD	GSIM	NGSIM
TID2008	SROCC	0.524	0.704	0.774	0.852	0.855	0.880	0.884	0.891	0.855	0.892
	KROCC	0.369	0.534	0.576	0.654	0.663	0.694	0.704	0.708	0.665	0.713
	PLCC	0.530	0.682	0.773	0.842	0.857	0.873	0.885	0.879	0.846	0.886
	RMSE	1.137	0.981	0.851	0.729	0.689	0.652	0.624	0.640	0.715	0.622
CSIQ	SROCC	0.805	0.810	0.875	0.913	0.921	0.924	0.932	0.957	0.912	0.962
	KROCC	0.608	0.624	0.690	0.739	0.752	0.756	0.768	0.813	0.740	0.825
	PLCC	0.800	0.800	0.861	0.899	0.914	0.912	0.922	0.954	0.897	0.961
	RMSE	0.157	0.157	0.133	0.114	0.106	0.100	0.101	0.079	0.115	0.073
LIVE	SROCC	0.875	0.927	0.947	0.944	0.956	0.963	0.962	0.960	0.955	0.950
	KROCC	0.686	0.761	0.796	0.792	0.817	0.833	0.839	0.823	0.813	0.815
	PLCC	0.872	0.923	0.944	0.943	0.952	0.959	0.953	0.960	0.943	0.946
	RMSE	13.36	10.50	8.944	9.095	8.347	7.678	7.003	7.62	9.037	7.455

下載: 導(dǎo)出CSV

表 2 10種不同IQA算法在TID2008，CSIQ, LIVE數(shù)據(jù)庫(kù)單一失真性能(SROCC)的比較

數(shù)據(jù)庫(kù)	失真類(lèi)型	PSNR	VSNR	SSIM	MS-SSIM	IW-SSIM	FSIM	ESSIM	GMSD	GSIM	NGSIM
TID2008	AWN	0.907	0.772	0.811	0.809	0.786	0.857	0.885	0.918	0.857	0.902
	ANMC	0.899	0.779	0.803	0.805	0.792	0.851	0.813	0.898	0.809	0.873
	SCN	0.917	0.766	0.815	0.819	0.771	0.848	0.913	0.913	0.890	0.929
	JPEG	0.872	0.917	0.925	0.934	0.918	0.928	0.943	0.952	0.939	0.956
	JP2K	0.813	0.951	0.962	0.973	0.973	0.977	0.975	0.980	0.975	0.958
	J2TE	0.831	0.790	0.858	0.852	0.820	0.854	0.879	0.883	0.892	0.926
CSIQ	AWGN	0.936	0.924	0.897	0.947	0.938	0.926	0.949	0.968	0.944	0.966
	JPEG	0.888	0.903	0.954	0.963	0.966	0.965	0.964	0.965	0.963	0.966
	JP2K	0.936	0.948	0.960	0.968	0.968	0.968	0.967	0.972	0.964	0.974
	FNIOSE	0.933	0.908	0.892	0.933	0.905	0.923	0.943	0.950	0.938	0.962
	BLUR	0.929	0.944	0.960	0.971	0.978	0.972	0.962	0.971	0.958	0.967
	CONTRST	0.862	0.870	0.792	0.952	0.953	0.942	0.939	0.904	0.950	0.946
LIVE	JPEG2	0.895	0.955	0.961	0.962	0.964	0.971	0.980	0.971	0.958	0.972
	JPEG	0.880	0.965	0.976	0.981	0.980	0.983	0.981	0.978	0.909	0.960
	AWGN	0.985	0.978	0.969	0.973	0.966	0.965	0.976	0.974	0.977	0.993
	BLUR	0.782	0.941	0.951	0.954	0.972	0.970	0.991	0.957	0.951	0.939
	FASTFA	0.890	0.902	0.955	0.947	0.944	0.949	0.947	0.942	0.939	0.956

下載: 導(dǎo)出CSV

參考文獻(xiàn)(22)

BAE S H and KIM M. A novel image quality assessment with globally and locally consilient visual quality perception[J]. IEEE Transactions on Image Processing, 2016, 25(5): 2392–2406. doi: 10.1109/TIP.2016.2545863

WANG Hanli, FU Jie, LIN Weisi, et al. Image quality assessment based on local linear information and distortion-specific compensation[J]. IEEE Transactions on Image Processing, 2017, 26(2): 915–926. doi: 10.1109/TIP.2016.2639451

DI E C and JACOVITTI G. A detail based method for linear full reference image quality prediction[J]. IEEE Transactions on Image Processing, 2017, 27(1): 179–192. doi: 10.1109/TIP.2017.2757139

CHANDLER D M and HEMAMI S S. VSNR: A wavelet-based visual signal-to-noise ratio for natural images[J]. IEEE Transactions on Image Processing, 2007, 16(9): 2284–2298. doi: 10.1109/TIP.2007.901820

褚江, 陳強(qiáng), 楊曦晨. 全參考圖像質(zhì)量評(píng)價(jià)綜述[J]. 計(jì)算機(jī)應(yīng)用研究, 2014, 31(1): 13–22. doi: 10.3969/j.issn.1001-3695.2014.01.003

CHU Jiang, CHEN Qiang, and YANG Xichen. Review on full reference image quality assessment algorithms[J]. Application Research of Computers, 2014, 31(1): 13–22. doi: 10.3969/j.issn.1001-3695.2014.01.003

WANG Zhou and BOVIK A C. Mean squared error: love it or leave it? A new look at signal fidelity measures[J]. IEEE Signal Processing Magazine, 2009, 26(1): 98–117. doi: 10.1109/MSP.2008.930649

HUYNH-THU Q and GHANBARI M. Scope of validity of PSNR in image/video quality assessment[J]. Electronics Letters, 2008, 44(13): 800–801. doi: 10.1049/el:20080522

WANG Zhou, BOVIK A C, SHEIKH H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13(4): 600–612. doi: 10.1109/TIP.2003.819861

WANG Zhou, SIMONCELLI E P, and BOVIK A C. Multiscale structural similarity for image quality assessment[C]. Proceedings of 37th IEEE Asilomar Conference on Signals, Systems and Computers, Pacific Grove, USA, 2003: 1398–1402.

LI Chaofeng and BOVIK A C. Three-component weighted structural similarity index[C]. SPIE Conference on Image Quality and System Performance, San Jose, USA, 2009, 7242: 72420Q–72420Q-9.

WANG Zhou and LI Qiang. Information content weighting for perceptual image quality assessment[J]. IEEE Transactions on Image Processing, 2011, 20(5): 1185–1198. doi: 10.1109/TIP.2010.2092435

ZHANG Lin, ZHANG Lei, MOU Xuanqin, et al. FSIM: A feature similarity index for image quality assessment[J]. IEEE Transactions on Image Processing, 2011, 20(8): 2378–2386. doi: 10.1109/TIP.2011.2109730

LIU Anmin, LIN Weisi, and NARWARIA M. Image quality assessment based on gradient similarity[J]. IEEE Transactions on Image Processing, 2012, 21(4): 1500–1512. doi: 10.1109/TIP.2011.2175935

XUE Wufeng, ZHANG Lei, MOU Xuanqin, et al. Gradient magnitude similarity deviation: A highly efficient perceptual image quality index[J]. IEEE Transactions on Image Processing, 2014, 23(2): 684–695. doi: 10.1109/TIP.2013.2293423

ZHANG Xuande, FENG Xiangchu, WANG Weiwei, et al. Edge strength similarity for image quality assessment[J]. IEEE Signal Processing Letters, 2013, 20(4): 319–322. doi: 10.1109/LSP.2013.2244081

WANG Tonghan, JIA Huizhen, and SHU Huazhong. Full-reference image quality assessment algorithm based on gradient magnitude and histogram of oriented gradient[J]. Journal of Southeast University, 2018, 48(2): 276–281. doi: 10.3969/j.issn.1001-0505.2018.02.014

NI Zhangkai, MA Lin, ZENG Huanqiang, et al. Gradient direction for screen content image quality assessment[J]. IEEE Signal Processing Letters, 2016, 23(10): 1394–1398. doi: 10.1109/LSP.2016.2599294

DING Li, HUANG Hua, and ZANG Yu. Image quality assessment using directional anisotropy structure measurement[J]. IEEE Transactions on Image Processing, 2017, 26(4): 1799–1809. doi: 10.1109/TIP.2017.2665972

張帆, 張偌雅, 李珍珍. 基于對(duì)稱(chēng)相位一致性的圖像質(zhì)量評(píng)價(jià)方法[J]. 激光與光電子學(xué)進(jìn)展, 2017, 54(10): 194–202. doi: 10.3788/LOP54.101003

ZHANG Fan, ZHANG Ruoya, and LI Zhenzhen. Image quality assessment based on symmetry phase congruency[J]. Laser &Optoelectronics Progress, 2017, 54(10): 194–202. doi: 10.3788/LOP54.101003

PONOMARENKO N, LUKIN V, ZELENSKY A, et al. TID2008: A database for evaluation of full-reference visual quality assessment metrics[OL]. http://www.ponomarenko.info/papers/mre2009tid.pdf. 2016.10.

LARSON EC and CHANDLER D. Categorical subjective image quality (CSIQ) database[OL]. http://vision.okstate.edu/csiq, 2016.10.

SHEIKH H R, WANG Zhou, BOVIK A C, et al. Image and video quality assessment research at LIVE[OL]. http://live.ece.utexas.edu/rese-arch/quality/. 2016.10.

相關(guān)文章

施引文獻(xiàn)

資源附件(0)

訪(fǎng)問(wèn)統(tǒng)計(jì)