Laser & Optoelectronics Progress, Volume. 62, Issue 2, 0237004(2025)
Stereo Image Quality Assessment Based on Top-Down Visual Mechanism
Figures & Tables(13)
Table 1. Performance comparison on LIVE 3D database
View tableTable 1. Performance comparison on LIVE 3D database
Method LIVE 3D Phase I LIVE 3D Phase II PLCC SROCC RMSE PLCC SROCC RMSE Ref. ［ 30 ］0.8950 0.8910 7.2470 0.8800 0.8800 5.1020 Ref. ［ 12 ］0.9470 0.9430 5.3360 Ref. ［ 15 ］0.9730 0.9650 3.6820 0.9570 0.9470 3.2700 Ref. ［ 16 ］0.9620 0.9500 0.9460 0.9380 Ref. ［ 17 ］0.9720 0.9620 0.9530 0.9510 Ref. ［ 18 ］0.9779 0.9656 2.6077 0.9717 0.9529 2.2771 Ref. ［ 21 ］0.9705 0.9725 4.0457 0.9624 0.9557 3.4598 Ref. ［ 22 ］0.9746 0.9716 3.5526 0.9671 0.9627 2.9853 Proposed method 0.9773 0.9735 3.9677 0.9674 0.9627 3.0062
Table 2. Performance comparison on WIVC 3D database
View tableTable 2. Performance comparison on WIVC 3D database
Method WIVC 3D Phase I WIVC 3D Phase II PLCC SROCC RMSE PLCC SROCC RMSE Ref. ［ 18 ］0.9690 0.9599 2.9508 0.9580 0.9501 3.0506 Ref. ［ 21 ］0.9525 0.9502 3.3431 0.9665 0.9524 4.2889 Ref. ［ 22 ］0.9739 0.9608 3.4537 0.9768 0.9765 2.8933 Proposed method 0.9742 0.9725 3.3066 0.9733 0.9740 3.5029
Table 3. PLCC comparison of individual distortion types on two LIVE 3D databases
View tableTable 3. PLCC comparison of individual distortion types on two LIVE 3D databases
Method LIVE 3D Phase I LIVE 3D Phase II BLUR FF JP2K JPEG WN BLUR FF JP2K JPEG WN Ref. ［ 30 ］0.9170 0.7350 0.9070 0.6950 0.9170 0.9410 0.9320 0.8990 0.9010 0.9470 Ref. ［ 12 ］0.9300 0.8830 0.9260 0.7400 0.9440 Ref. ［ 15 ］0.9740 0.9650 0.9880 0.9160 0.9880 0.9550 0.9940 0.9050 0.9330 0.9720 Ref. ［ 17 ］0.9880 0.9390 0.9840 0.9060 0.9470 0.9880 0.9640 0.9560 0.8250 0.9540 Ref. ［ 21 ］0.9715 0.9221 0.9916 0.9406 0.9601 0.9987 0.9609 0.9712 0.8754 0.9637 Ref. ［ 22 ］0.9928 0.9425 0.9876 0.9173 0.9724 0.9969 0.9686 0.9810 0.9046 0.9650 Proposed method 0.9958 0.9590 0.9638 0.9366 0.9959 0.9990 0.9651 0.9324 0.9053 0.9960
Table 4. SROCC comparison of individual distortion types on two LIVE 3D databases
View tableTable 4. SROCC comparison of individual distortion types on two LIVE 3D databases
Method LIVE 3D Phase I LIVE 3D Phase II BLUR FF JP2K JPEG WN BLUR FF JP2K JPEG WN Ref. ［ 30 ］0.8780 0.6520 0.8630 0.6170 0.9190 0.9000 0.9330 0.8670 0.8670 0.9500 Ref. ［ 12 ］0.9090 0.8340 0.9310 0.6930 0.9460 Ref. ［ 15 ］0.8550 0.9170 0.9610 0.9120 0.9650 0.6000 0.9510 0.8740 0.7470 0.9420 Ref. ［ 17 ］0.9450 0.9000 0.9650 0.8790 0.9210 0.9510 0.9690 0.9540 0.8160 0.9230 Ref. ［ 21 ］0.9667 0.9473 0.9780 0.9393 0.9455 0.9231 0.9636 0.9525 0.8555 0.9341 Ref. ［ 22 ］0.9833 0.9604 0.9604 0.9036 0.9701 0.8811 0.9455 0.9608 0.8658 0.9341 Proposed method 0.9903 0.9176 0.9547 0.9393 0.9966 0.9771 0.9588 0.9439 0.8674 0.9762
Table 5. Performance comparison of cross-dataset validation on two LIVE 3D databases
View tableTable 5. Performance comparison of cross-dataset validation on two LIVE 3D databases
Method LIVE 3D Phase I/II LIVE 3D Phase II/I PLCC SROCC PLCC SROCC Ref. ［ 15 ］0.7100 0.9320 Ref. ［ 17 ］0.8480 0.8330 0.9150 0.9210 Ref. ［ 18 ］0.8595 0.7972 0.9184 0.9149 Ref. ［ 21 ］0.8471 0.8074 0.9101 0.9061 Ref. ［ 22 ］0.8212 0.7625 0.9052 0.9133 Proposed method 0.8573 0.8480 0.9226 0.9209
Table 6. Performance comparison of cross-dataset validation on two WIVC 3D databases
View tableTable 6. Performance comparison of cross-dataset validation on two WIVC 3D databases
Method WIVC 3D Phase I / II WIVC 3D Phase II / I PLCC SROCC PLCC SROCC Ref. ［ 18 ］0.7816 0.7502 0.8981 0.8712 Ref. ［ 21 ］0.9223 0.9117 0.9364 0.9490 Ref. ［ 22 ］0.9171 0.9023 0.9309 0.9274 Proposed method 0.9150 0.9088 0.9323 0.9289
Table 7. Performance comparison of SAT ablation study on two LIVE 3D databases
View tableTable 7. Performance comparison of SAT ablation study on two LIVE 3D databases
Scheme Param. /106 LIVE 3D Phase I LIVE 3D Phase II PLCC SROCC PLCC SROCC No attention w/o SAT 7.07 0.9415 0.9404 0.9345 0.9304 Attention in monocular M-SE 7.88 0.9534 0.9530 0.9515 0.9510 M-CBAM 7.88 0.9579 0.9538 0.9566 0.9565 M-GC 7.88 0.9554 0.9540 0.9522 0.9522 Attention in binocular B-SE 7.47 0.9536 0.9520 0.9515 0.9511 B-CBAM 7.47 0.9588 0.9540 0.9567 0.9563 B-GC 7.47 0.9552 0.9544 0.9525 0.9511 SAT B2B 7.47 0.9654 0.9653 0.9639 0.9600 B2M 7.45 0.9671 0.9660 0.9658 0.9611 B2B+B2M 7.47 0.9773 0.9735 0.9674 0.9627
Table 8. Performance comparison of
ablation study on two LIVE 3D databases View tableTable 8. Performance comparison of
ablation study on two LIVE 3D databases Method LIVE 3D Phase I LIVE 3D Phase II PLCC SROCC PLCC SROCC 0.9454 0.9475 0.9405 0.9407 0.9576 0.9572 0.9546 0.9534 0.9579 0.9572 0.9529 0.9513 w/o （ ） 0.9590 0.9573 0.9523 0.9515 Proposed method 0.9773 0.9735 0.9674 0.9627
Table 9. Performance comparison of different pooling strategies on two LIVE 3D databases
View tableTable 9. Performance comparison of different pooling strategies on two LIVE 3D databases
Pooling strategy LIVE 3D Phase I LIVE 3D Phase II Avg Max Min PLCC SROCC PLCC SROCC + - 0.9527 0.9503 0.9468 0.9430 + - 0.9676 0.9653 0.9589 0.9584 + - 0.9599 0.9582 0.9497 0.9486 + - 0.9642 0.9614 0.9587 0.9569 - + 0.9773 0.9735 0.9674 0.9627
Table 10. Complexity comparison on the LIVE 3D Phase I database
View tableTable 10. Complexity comparison on the LIVE 3D Phase I database
Method Param. /106 Inference time /s Ref. ［ 18 ］3830.00 Ref. ［ 21 ］28.57 0.0963 Ref. ［ 22 ］35.18 0.1712 Proposed method 7.47 0.0075
Tools
Get Citation
Copy Citation Text
Sumei Li, Huilin Zhang. Stereo Image Quality Assessment Based on Top-Down Visual Mechanism[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0237004
Paper Information
Category: Digital Image Processing
Received: May. 7, 2024
Accepted: May. 16, 2024
Published Online: Jan. 3, 2025
The Author Email:
CSTR:32186.14.LOP241231
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20