Non-Reference Blur Image Quality Evaluation Based on Saliency Object Classification

Feipeng Shen; Tong Zhu; Henan Zhang; Zhenghao Chen

doi:10.3788/LOP202158.2210015

Laser & Optoelectronics Progress, Volume. 58, Issue 22, 2210015(2021)

Non-Reference Blur Image Quality Evaluation Based on Saliency Object Classification

Feipeng Shen^1、*, Tong Zhu¹, Henan Zhang^1,2, and Zhenghao Chen¹

¹School of Information and Control Engineering, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China

²Engineering Research Center of Intelligent Control for Underground Space, Ministry of Education, China University of Mining and Technology, Xuzhou, Jiangsu 221116

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. Two images with different content. (a) Saliency object image; (b) no-saliency object image

Download full size

Fig. 2. Framework of quality evaluation method proposed in this paper

Download full size

Fig. 3. Framework diagram of SDOC algorithm

Download full size

Fig. 4. Example images of saliency region segmentation. (a) Example image with saliency object; (b) first-level saliency region segmentation map of Fig. 4(a); (c) second-level saliency region segmentation map of Fig. 4(a); (d) example image without object; (e) first-level saliency region segmentation map of Fig. 4(d); (f) second-level saliency region segmentation map of Fig. 4(d)

Download full size

Fig. 5. Scatter plot of BLUR database classification results

Download full size

Fig. 6. Test results of global blur features in each database. (a) LIVE database; (b) CSIQ database; (c) TID2013 database; (d) BLUR database

Download full size

Fig. 7. Test results of SOA algorithm in each database for different β. (a) PLCC; (b) SROCC

Download full size

Fig. 8. Scatter plots of results of SOA algorithm for each database and corresponding fitting curves. (a) LIVE; (b) CSIQ; (c) TID2013; (d) BLUR

Download full size

Fig. 9. Comparison of prediction curves of various algorithms. (a) CPBD; (b) LPC; (c) GMVG; (d) SOA

Download full size

Table 1. Introduction of parameters of each database
View table
Table 1. Introduction of parameters of each database
Database Distortion type Number of samples Resolution Subjective evaluation Range
LIVE Gaussian blur 174 768×512 DMOS [0, 100]
CSIQ 150 512×512 DMOS [0, 1]
TID2013 125 512×384 MOS [0, 9]
BLUR 144 481×321 MOS [0, 10]

Table 2. LIVE database test results
View table
Table 2. LIVE database test results
Algorithm PLCC SROCC KROCC
JNB 0.8221 0.8419 0.6651
CPBD 0.9123 0.9429 0.8043
LPC 0.8172 0.9594 0.8241
NRIQAVR 0.8023 0.8441 0.7060
GMVG 0.9335 0.9633 0.8377
SOA 0.9667 0.9622 0.8369

Table 3. CSIQ database test results
View table
Table 3. CSIQ database test results
Algorithm PLCC SROCC KROCC
JNB 0.2508 0.7624 0.5971
CPBD 0.8292 0.8846 0.7081
LPC 0.9096 0.9068 0.7197
NRIQAVR 0.9152 0.8874 0.7688
GMVG 0.9264 0.9500 0.8153
SOA 0.9421 0.9358 0.7892

Table 4. TID2013 database test results
View table
Table 4. TID2013 database test results
Algorithm PLCC SROCC KROCC
JNB 0.6876 0.6902 0.5137
CPBD 0.8491 0.8520 0.6470
LPC 0.8382 0.8888 0.6839
NRIQAVR 0.7819 0.7967 0.6533
GMVG 0.9125 0.9274 0.7600
SOA 0.9130 0.9185 0.7430

Table 5. BLUR database test results
View table
Table 5. BLUR database test results
Algorithm PLCC SROCC KROCC
JNB 0.6295 0.6691 0.7621
CPBD 0.8944 0.9077 0.8318
LPC 0.9108 0.9049 0.7058
NRIQAVR 0.8261 0.8448 0.6876
GMVG 0.7918 0.9000 0.7441
SOA 0.9173 0.9348 0.7786

Table 6. Comprehensive performance error values of different algorithms for each database
View table
Table 6. Comprehensive performance error values of different algorithms for each database
Algorithm LIVE CSIQ TID2013 BLUR
JNB 16.800 49.340 31.110 35.070
CPBD 7.240 14.310 14.945 9.895
LPC 11.170 9.180 13.650 9.215
NRIQAVR 17.680 9.870 21.070 16.455
GMVG 5.160 6.180 8.005 15.410
SOA 3.555 6.105 8.425 7.395

Table 7. F test related parameters of each database
View table
Table 7. F test related parameters of each database
Database Confidence level Freedom F_critical
LIVE 0.9 174 1.2150
CSIQ 150 1.2335
TID2013 125 1.2586
BLUR 144 1.2388

Table 8. Performance statistical results of SOA algorithm for each database
View table
Table 8. Performance statistical results of SOA algorithm for each database
Algorithm LIVE CSIQ TID2013 BLUR
JNB 1 1 1 1
CPBD 1 1 1 1
LPC 1 1 1 1
NRIQAVR 1 1 1 1
GMVG 1 0 0 1

Tools

Get Citation

Copy Citation Text

Feipeng Shen, Tong Zhu, Henan Zhang, Zhenghao Chen. Non-Reference Blur Image Quality Evaluation Based on Saliency Object Classification[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2210015

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites