Fig. 1. Framework of environmental prejudgment model for meter reading
Fig. 2. Meter images and DCT transform images under different fog concentrations. (a) Images with score of 1; (b) images with score of 3; (c) images with score of 5; (d) images with score of 7
Fig. 3. Meter images and LBP images after normalization pretreatment under different fog concentrations. (a) Score of 1; (b) score of 4; (c) score of 7
Fig. 4. Meter images and depth images at different fog concentrations. (a) Score of 1; (b) score of 4; (c) score of 7
Fig. 5. Partial meter images in dataset
Fig. 6. Final evaluation scatterplots of different algorithms. (a) GM-LOG-BIQA algorithm; (b) BRISQUE algorithm; (c) GWH-GLBP-BIQA algorithm; (d) BLINDS2 algorithm; (e) SSEQ algorithm; (f) proposed algorithm
Fig. 7. Meter images and their corresponding depth estimation histograms at different distances. (a) 1.00 m; (b) 1.25 m; (c) 1.50 m; (d) 1.75 m
Algorithm | Type | SROCC | PLCC | RMSE |
---|
LPSI | I | 0.5224 | 0.5369 | 2.1758 | NIQE | I | 0.7508 | 0.6940 | 1.7389 | BIQI | I | 0.8239 | 0.8277 | 1.4612 | ASIQE | I | 0.8396 | 0.8657 | 2.9554 | GM-LOG-BIQA | II | 0.8347 | 0.8176 | 1.3644 | BRISQUE | II | 0.8478 | 0.8486 | 1.3606 | GWH-GLBP-BIQA | II | 0.8676 | 0.8482 | 1.2479 | BLINDS2 | II | 0.9168 | 0.8817 | 0.9887 | SSEQ | II | 0.9178 | 0.8852 | 0.9890 | Proposed algorithm | II | 0.9493 | 0.9353 | 0.8280 |
|
Table 1. Performance comparison of different algorithms
Algorithm | Without depth feature | With depth feature |
---|
SROCC | PLCC | RMSE | SROCC | PLCC | RMSE |
---|
BLINDS2 | 0.8202 | 0.7907 | 1.2734 | 0.9168 | 0.8817 | 0.9887 | BRISQUE | 0.7614 | 0.7441 | 1.4170 | 0.8478 | 0.8486 | 1.3606 | GM-LOG-BIQA | 0.7211 | 0.7153 | 1.6692 | 0.8347 | 0.8176 | 1.3644 | GWH-GLBP-BIQA | 0.7919 | 0.7541 | 1.5858 | 0.8676 | 0.8482 | 1.2479 | SSEQ | 0.8275 | 0.8026 | 1.2948 | 0.9178 | 0.8852 | 0.9890 | Proposed algorithm | 0.9107 | 0.8820 | 1.4994 | 0.9493 | 0.9353 | 0.8280 |
|
Table 2. Comparison of performance of different algorithms with or without depth map feature
Feature | SROCC | PLCC | RMSE |
---|
Without DCT feature | 0.8520 | 0.8467 | 1.9713 | Without structural feature | 0.8771 | 0.8567 | 1.9711 | Without all features | 0.9493 | 0.9353 | 0.8280 |
|
Table 3. Comparison of performance of proposed algorithm after removing DCT features and spatial structure features
β | SROCC | PLCC | RMSE |
---|
2.0 | 0.9327 | 0.9033 | 1.0548 | 2.5 | 0.9356 | 0.9168 | 0.9877 | 3.0 | 0.9350 | 0.9142 | 1.0127 | Fixed value | 0.9493 | 0.9353 | 0.8280 |
|
Table 4. Performance comparison with different β values