Author Affiliations
1Engineering & Technical College of Chengdu University of Technology, Leshan, Sichuan 614000, China;2School of Information and Control Engineering, China University of Mining and Technology, Xuzhou, Jiangsu 221116, Chinashow less
Fig. 1. Framework of our algorithm
Fig. 2. Original image and gamut mapping images with decreasing quality. (a) Original image; (b) gamut mapping image1; (c) gamut mapping image2; (d) gamut mapping image3
Fig. 3. Changes in the frequency domain moment and entropy of the image with frequency. (a) Change of image frequency domain entropy with frequency; (b) change of image frequency domain mean with frequency; (c) change of image frequency domain standard deviation with frequency
Fig. 4. Empirical histograms of relative chroma
Fig. 5. Empirical histograms of relative hue
Fig. 6. Empirical histogram of relative hue and chroma
Fig. 7. Structure of the AdaBoosting BPNN; (a) Structure of the AdaBoosting algorithm; (b) structure of BPNN
Fig. 8. Performance comparison of grayscale features and color features. (a) BS database; (b) IG database; (c) LC database
Fig. 9. Influence of peak value features on algorithm performance. (a) BS database; (b) IG database; (c) LC database
Fig. 10. Example of gamut mapping images. (a) MOS is 0.6268; (b) MOS is 0.5972; (c) MOS is 0.2927; (d) MOS is 0.1341
Database | Referenceimage | Distortedimage | Evaluation | GAM |
---|
BS | 97 | 1067 | 5199 | 11 | IG | 65 | 520 | 3698 | 8 | LC | 72 | 576 | 5209 | 8 |
|
Table 1. Gamut mapping image quality evaluation databases
Algorithm | BS | IG | LC |
---|
PLCC | SRCC | KRCC | RMSE | PLCC | SRCC | KRCC | RMSE | PLCC | SRCC | KRCC | RMSE |
---|
BRISQUE | 0.7633 | 0.5678 | 0.4126 | 0.4386 | 0.5153 | 0.4654 | 0.3345 | 0.4739 | 0.5026 | 0.5274 | 0.3802 | 0.4229 | BIQI | 0.6188 | 0.4422 | 0.3135 | 0.5335 | 0.3680 | 0.3078 | 0.2227 | 0.5115 | 0.3777 | 0.3521 | 0.2486 | 0.4516 | DESIQUE | 0.8213 | 0.5941 | 0.4354 | 0.3878 | 0.5987 | 0.5666 | 0.4211 | 0.4440 | 0.5692 | 0.5973 | 0.4429 | 0.4367 | DIIVINE | 0.7339 | 0.5457 | 0.3949 | 0.4603 | 0.4289 | 0.3694 | 0.2694 | 0.4986 | 0.4210 | 0.4111 | 0.2920 | 0.4441 | NFERM | 0.7441 | 0.5556 | 0.4072 | 0.4539 | 0.4399 | 0.4150 | 0.2968 | 0.4942 | 0.4934 | 0.4985 | 0.3617 | 0.4263 | BLIINDS_II | 0.7081 | 0.5499 | 0.4031 | 0.4777 | 0.3646 | 0.3102 | 0.2184 | 0.5127 | 0.4274 | 0.3323 | 0.2370 | 0.4449 | IDEAL | 0.7859 | 0.6652 | 0.4994 | 0.4173 | 0.6195 | 0.6139 | 0.4550 | 0.4327 | 0.5780 | 0.5989 | 0.4417 | 0.3977 | IL_NIQE | 0.5545 | 0.4849 | 0.3937 | 0.4923 | 0.3560 | 0.3416 | 0.2808 | 0.4019 | 0.4748 | 0.3459 | 0.3439 | 0.3842 | NIQE | 0.5840 | 0.4479 | 0.3840 | 0.5132 | 0.3724 | 0.3667 | 0.2936 | 0.4675 | 0.4748 | 0.3247 | 0.2418 | 0.3458 | Ours | 0.8170 | 0.6774 | 0.5100 | 0.3918 | 0.7369 | 0.7086 | 0.5526 | 0.3773 | 0.6256 | 0.6154 | 0.4630 | 0.3849 |
|
Table 2. Performance comparison of different algorithms in three databases
Image | MOS | DESIQUE | BRISQUE | IL_NIQE | IDEAL | Ours |
---|
Fig.10(a) | 0.6268 | 21.2382 | 18.4999 | 20.5977 | 4.6353 | 0.6736 | Fig.10(b) | 0.5927 | 21.4467 | 20.2568 | 22.5301 | 4.7215 | 0.6213 | Fig.10(c) | 0.2927 | 22.1088 | 20.7486 | 22.2513 | 4.3958 | 0.3042 | Fig.10(d) | 0.1341 | 20.5985 | 26.2595 | 22.3621 | 4.1315 | 0.1556 |
|
Table 3. Predicted results by different algorithms