Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 57, Issue 18, 181509(2020)

No Reference Video Quality Assessment Based on Spatio-Temporal Features and Attention Mechanism

Ze Zhu1, Qingbing Sang1,2、*, and Hao Zhang1
Author Affiliations
  • 1School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
  • 2Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence, Wuxi, Jiangsu 214122, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (17)
    Equations (0)
    References (25)
    Cited By (1)
    Tools
    Paper Information
    Topics
    Figures & Tables(17)
    Network structure

    Fig. 1. Network structure

    Download full size
    Schematic of GRU network structure

    Fig. 2. Schematic of GRU network structure

    Download full size
    Attention model

    Fig. 3. Attention model

    Download full size
    1st frame of different distorted videos. (a) Riverbed; (b) sunflower; (c) station; (d) tractor

    Fig. 4. 1st frame of different distorted videos. (a) Riverbed; (b) sunflower; (c) station; (d) tractor

    Download full size
    Flow chart of video data processing

    Fig. 5. Flow chart of video data processing

    Download full size
    Scatter plot of prediction results on LIVE video library

    Fig. 6. Scatter plot of prediction results on LIVE video library

    Download full size
    Relationship curves between number of training sets of different proportions and evaluation results

    Fig. 7. Relationship curves between number of training sets of different proportions and evaluation results

    Download full size
    Scatter plot of prediction results on CSIQ video library

    Fig. 8. Scatter plot of prediction results on CSIQ video library

    Download full size
    Scatter plot of prediction results on IVP video library

    Fig. 9. Scatter plot of prediction results on IVP video library

    Download full size

    • Table 1. Network parameter setting

      View table

      Table 1. Network parameter setting

      Layer nameOutput sizeParameter
      Conv1,Conv224000×48×64Size: 3×3; filters: 64
      Max pooling112000×24×64Size: 2×2; stride: 2×2
      Conv3,Conv412000×24×128Size: 3×3; filters: 128
      Max pooling26000×12×128Size: 2×2; stride: 2×2
      Conv5,Conv6,Conv76000×12×256Size: 3×3; filters: 256
      Max pooling33000×6×256Size: 2×2; stride: 2×2
      Conv8,Conv9,Conv103000×6×512Size: 3×3; filters: 512
      Max pooling41500×3×512Size: 2×2; stride: 2×2
      Conv11,Conv12,Conv131500×3×512Size: 3×3; filters: 512
      Max pooling5749×1×512Size: 2×2; stride: 3×3
      GRU1×1×512512
      Attention1×512/
      FC1×11

    • Table 2. Performance comparison of different algorithms on LIVE video library

      View table

      Table 2. Performance comparison of different algorithms on LIVE video library

      AlgorithmSROCCPLCC
      PSNR[23]0.53980.5645
      SSIM[24]0.73640.7470
      ST-MAD[6]0.82510.8332
      STRRED[25]0.80070.8119
      FS-MOVIE[7]0.84820.8636
      V-BLIINDS[4]0.83770.8471
      Ours without attention0.85570.8633
      Ours with attention0.87980.8910

    • Table 3. Comparison of SROCC values of different algorithms for single distortion type

      View table

      Table 3. Comparison of SROCC values of different algorithms for single distortion type

      AlgorithmWirelessIPH.264MPEG-2
      PSNR[23]0.65740.41670.45850.3862
      SSIM[24]0.72890.65340.73130.6684
      ST-MAD[6]0.80990.77580.90210.8461
      STRRED[25]0.78570.77220.81930.7193
      FS-MOVIE[7]0.81390.77220.84900.8609
      V-BLIINDS[4]0.84550.78980.85870.8377
      Ours withoutattention0.84870.83160.84680.8331
      Ours withattention0.86170.84580.85850.8547

    • Table 4. Comparison of PLCC values of different algorithms for single distortion type

      View table

      Table 4. Comparison of PLCC values of different algorithms for single distortion type

      AlgorithmWirelessIPH.264MPEG-2
      PSNR[23]0.70580.47670.57460.3986
      SSIM[24]0.71840.77640.74200.6222
      ST-MAD[6]0.85910.80650.87960.8560
      STRRED[25]0.80530.85270.81410.7570
      FS-MOVIE[7]0.85990.80090.87650.8721
      V-BLIINDS[4]0.91340.90200.90380.8699
      Ours withoutattention0.90690.90990.87660.8745
      Ours withattention0.92030.91770.89620.8858

    • Table 5. Comparison of final evaluation results on LIVE video library

      View table

      Table 5. Comparison of final evaluation results on LIVE video library

      AlgorithmLiveData1LiveData2LiveData3LiveData4Average
      SROCCPLCCSROCCPLCCSROCCPLCCSROCCPLCCSROCCPLCC
      Ours withoutattention0.84780.87520.84820.86720.85440.84610.87240.86480.85570.8633
      Ours with attention0.86930.89080.89100.90040.88520.87580.87350.89690.87980.8910

    • Table 6. Comparison of running time of different methods on “Tractor” video

      View table

      Table 6. Comparison of running time of different methods on “Tractor” video

      AlgorithmTime /s
      PSNR[23]3.09
      SSIM[24]11.34
      ST-MAD[6]335.90
      STRRED[25]54.94
      FS-MOVIE[7]4444.20
      Ours with attention1291.20

    • Table 7. Performance comparison of different algorithms on CSIQ video library

      View table

      Table 7. Performance comparison of different algorithms on CSIQ video library

      AlgorithmSROCCPLCC
      PSNR[23]0.72530.7932
      SSIM[24]0.86610.8517
      ST-MAD[6]0.81740.8266
      STRRED[25]0.88220.8734
      FS-MOVIE[7]0.80670.8053
      V-BLIINDS[4]0.83510.8449
      Ours with attention0.89090.8991

    • Table 8. Performance comparison of different algorithms on IVP video library

      View table

      Table 8. Performance comparison of different algorithms on IVP video library

      AlgorithmSROCCPLCC
      PSNR[23]0.70640.7299
      SSIM[24]0.76940.7667
      ST-MAD[6]0.82350.8284
      STRRED[25]0.87610.8853
      FS-MOVIE[7]0.81770.8359
      V-BLIINDS[4]0.85520.8441
      Ours with attention0.90640.9135
    Tools

    Get Citation

    Copy Citation Text

    Ze Zhu, Qingbing Sang, Hao Zhang. No Reference Video Quality Assessment Based on Spatio-Temporal Features and Attention Mechanism[J]. Laser & Optoelectronics Progress, 2020, 57(18): 181509

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Machine Vision

    Received: Jan. 7, 2020

    Accepted: Feb. 24, 2020

    Published Online: Sep. 2, 2020

    The Author Email: Qingbing Sang (sangqb@163.com)

    DOI:10.3788/LOP57.181509

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology