• Laser & Optoelectronics Progress
  • Vol. 58, Issue 18, 1811014 (2021)
Yingqian Wang, Longguang Wang, Zhengyu Liang, Wei An, and Jungang Yang*
Author Affiliations
  • College of Electronic Science and Technology, National University of Defense Technology, Changsha, Hunan 410073, China
  • show less
    DOI: 10.3788/LOP202158.1811014 Cite this Article Set citation alerts
    Yingqian Wang, Longguang Wang, Zhengyu Liang, Wei An, Jungang Yang. Stereo Image Super-Resolution: A Survey[J]. Laser & Optoelectronics Progress, 2021, 58(18): 1811014 Copy Citation Text show less
    Examples of commercial stereo cameras
    Fig. 1. Examples of commercial stereo cameras
    Stereo camera and stereo images. (a) Schematic diagram of imaging model of stereo camera; (b) images recorded by stereo camera in Fig. 2 (a); (c) real scene images recorded by stereo camera taken from KITTI 2015 dataset[21]
    Fig. 2. Stereo camera and stereo images. (a) Schematic diagram of imaging model of stereo camera; (b) images recorded by stereo camera in Fig. 2 (a); (c) real scene images recorded by stereo camera taken from KITTI 2015 dataset[21]
    Epipolar constraint relation of stereo images
    Fig. 3. Epipolar constraint relation of stereo images
    Sample images in different stereo image datasets[33]
    Fig. 4. Sample images in different stereo image datasets[33]
    Comparison of visual effects of different algorithms at 2× magnification
    Fig. 5. Comparison of visual effects of different algorithms at 2× magnification
    Comparison of visual effects of different algorithms at 4× magnification
    Fig. 6. Comparison of visual effects of different algorithms at 4× magnification
    PSNR values achieved by PASSRnet algorithm in different testing sets at 4× magnification . (a) Testing set in KITTI 2015; (b) testing set in Middlebury; (c) testing set in Flickr1024; (d) testing set in ETH3D
    Fig. 7. PSNR values achieved by PASSRnet algorithm in different testing sets at 4× magnification . (a) Testing set in KITTI 2015; (b) testing set in Middlebury; (c) testing set in Flickr1024; (d) testing set in ETH3D
    AlgorithmMagnificationNumber of parameters /106KITTI 2012KITTI 2015MiddleburyFlickr1024
    PSNR/SSIMPSNR/SSIMPSNR/SSIMPSNR/SSIM
    Bicubic28.51/0.884228.61/0.897330.60/0.899024.94/0.8186
    VDSR[43]0.6630.30/0.908929.78/0.915032.77/0.910225.60/0.8534
    EDSR[44]38.630.96/0.922830.73/0.933534.95/0.949228.66/0.9087
    RDN[45]22.030.94/0.922730.70/0.933034.94/0.949128.64/0.9084
    RCAN[46]15.331.02/0.923230.77/0.933634.90/0.948628.63/0.9082
    StereoSR[22]1.0829.51/0.907329.33/0.916833.23/0.934825.96/0.8599
    PASSRnet[28]1.3730.81/0.919030.60/0.930034.23/0.942228.38/0.9038
    BSSRnet[31]1.8931.03/0.924130.74/0.934434.74/0.947528.53/0.9090
    iPASSR[32]1.3731.11/0.924030.81/0.934034.51/0.945428.60/0.9097
    SSRDE-FNet[26]2.1031.23/0.925430.90/0.935235.09/0.951128.85/0.9132
    Bicubic24.58/0.737224.38/0.734026.40/0.757221.82/0.6293
    VDSR[43]0.6625.60/0.772225.32/0.770327.69/0.794122.46/0.6718
    EDSR[44]38.926.35/0.801526.04/0.803929.23/0.839723.46/0.7285
    RDN[45]22.026.32/0.801426.04/0.804329.27/0.840423.47/0.7295
    RCAN[46]15.426.44/0.802926.22/0.806829.30/0.839723.48/0.7286
    StereoSR[22]1.0824.53/0.755524.21/0.751127.64/0.802221.70/0.6460
    PASSRnet[28]1.4226.34/0.798126.08/0.800228.72/0.823623.31/0.7195
    SRResNet+SAM[30]1.7326.44/0.801826.22/0.805428.83/0.829023.27/0.7233
    BSSRnet[31]1.9126.47/0.804926.17/0.807529.08/0.836223.40/0.7289
    iPASSR[32]1.4226.56/0.805326.32/0.808429.16/0.836723.44/0.7287
    SSRDE-FNet[26]2.2426.70/0.808226.43/0.811829.38/0.841123.59/0.7352
    Table 1. Numerical results achieved by different algorithms
    Training setTesting setin KITTI 2015Testing setin MiddleburyTesting set in Flickr1024Testing set in ETH3D
    PSNR/SSIMPSNR/SSIMPSNR/SSIMPSNR/SSIM
    Training set in KITTI 201524.28/0.74126.27/0.74921.77/0.61729.63/0.831
    Training set in Middlebury23.64/0.74326.62/0.77321.64/0.64628.66/0.843
    Training set in Flickr102425.08/0.77927.85/0.80722.64/0.69230.55/0.860
    Table 2. Numerical results achieved by StereoSR algorithm trained on different training sets for 60 epochs at 4× magnification
    Training setTesting setin KITTI 2015Testing setin MiddleburyTesting set in Flickr1024Testing set in ETH3D
    PSNR/SSIMPSNR/SSIMPSNR/SSIMPSNR/SSIM
    Training set in KITTI 201523.13/0.70325.42/0.76221.31/0.60026.95/0.789
    Training set in Middlebury25.18/0.77428.08/0.85322.54/0.67631.39/0.864
    Training set in Flickr102425.62/0.79128.69/0.87323.25/0.71831.94/0.877
    Table 3. Numerical results achieved by PASSRnet algorithm trained on different training sets for 60 epochs at 4× magnification
    Yingqian Wang, Longguang Wang, Zhengyu Liang, Wei An, Jungang Yang. Stereo Image Super-Resolution: A Survey[J]. Laser & Optoelectronics Progress, 2021, 58(18): 1811014
    Download Citation