Author Affiliations
Faculty of Information Science and Engineering, Ningbo University, Ningbo, Zhejiang 315211, Chinashow less
Fig. 1. Left images of the original satellite stereo images in the database
Fig. 2. Block diagram of building detection
Fig. 3. Original left image of building detection. (a) Corner detection results of original image; (b) DSM corresponding to Fig. (a); (c) building detection results
Fig. 4. Building detection in some areas. (a) Corner detection results; (b) DSM corresponding to Fig. (a); (c) building test results
Fig. 5. Building detection results. (a) Original image; (b) blur distortion image; (c) noise distortion image; (d) building detection of original image; (e) building detection result after blur distortion; (f) building detection result after noise distortion
Fig. 6. Building detection results. (a) Building detection results of original image; (b) building detection results after blur distortion; (c) building detection results after noise distortion
Fig. 7. Histogram of detection accuracy in the database
Fig. 8. Block diagram of objective quality evaluation. (a) Feature extraction; (b) sparse representation-based similarity measure
Fig. 9. SIFT features extraction. (a) SIFT features extraction of the original image; (b) SIFT features extraction of distorted image
Fig. 10. Scatter plots of evaluation prediction values and detection accuracy rates obtained by different evaluation methods. (a) MS-SSIM; (b) SSIM; (c) IFC; (d) VIF; (e) model in Ref.[27]; (f) FSIM; (g) GSM; (h) model in Ref.[29]; (i) proposed method
Criteria | | | | | Qf |
---|
PLCC | 0.8213 | 0.7203 | 0.8204 | 0.8469 | 0.9013 | SROCC | 0.7951 | 0.7048 | 0.8024 | 0.8024 | 0.8772 | KROCC | 0.5976 | 0.5279 | 0.5755 | 0.6065 | 0.7043 | RMSE | 6.6405 | 8.2312 | 6.7287 | 6.3688 | 5.0103 |
|
Table 1. Performance comparison of individual quality values
Model | PLCC | SROCC | KROCC | RMSE |
---|
MS-SSIM | 0.8146 | 0.7938 | 0.6288 | 6.9891 | SSIM | 0.6551 | 0.6135 | 0.4311 | 8.8133 | IFC | 0.8285 | 0.8154 | 0.6705 | 6.6789 | VIF | 0.8890 | 0.8722 | 0.6839 | 6.3872 | Model in Ref.[27] | 0.7929 | 0.7445 | 0.5477 | 7.0585 | FSIM | 0.6971 | 0.6623 | 0.4735 | 8.2684 | Model in Ref.[29] | 0.6891 | 0.6750 | 0.4830 | 8.4530 | GSM | 0.6610 | 0.6916 | 0.4987 | 8.6132 | Proposed | 0.9013 | 0.8772 | 0.7043 | 5.0103 |
|
Table 2. Comparison of overall performance of different evaluation methods
Distortion | Model |
---|
MS-SSIM | SSIM | IFC | VIF | Model in Ref.[27] | FSIM | Model in Ref.[29] | GSM | Proposed |
---|
Gblur | 0.8624 | 0.6651 | 0.8342 | 0.9152 | 0.8374 | 0.6494 | 0.6475 | 0.7013 | 0.8901 | WN | 0.8617 | 0.5321 | 0.8360 | 0.8823 | 0.7632 | 0.5836 | 0.5952 | 0.5579 | 0.9272 |
|
Table 3. PLCC values of different distortion types
Distortion | Model |
---|
MS-SSIM | SSIM | IFC | VIF | Model in Ref.[27] | FSIM | Model in Ref.[29] | GSM | Proposed |
---|
Gblur | 0.9384 | 0.7079 | 0.8714 | 0.8629 | 0.7774 | 0.7074 | 0.7147 | 0.7384 | 0.8567 | WN | 0.8967 | 0.6719 | 0.8463 | 0.8513 | 0.7577 | 0.6824 | 0.6904 | 0.7013 | 0.9076 |
|
Table 4. SROCC values of different distortion types