Author Affiliations
1School of Communication and Information Engineering, Xi'an University of Post and Telecommunications, Xi'an, Shaanxi 710121, China2School of Electronic Engineering, Xi'an University of Post and Telecommunications, Xi'an, Shaanxi 710121, China3School of Science, Xi'an University of Post and Telecommunications, Xi'an, Shaanxi 710121, Chinashow less
Fig. 1. Original images and spectra of images from R、G、B channels obtained by quaternion Fourier transform. (a)(e) Original images; (b)(f) images from R channel; (c)(g) images from G channel; (d)(h) images from B channel
Fig. 2. Monarch Gaussian blurred images under different standard deviations. (a) σ=0.5; (b) σ=1.0; (c) σ=1.5; (d) σ=2.0; (e) σ=2.5; (f) σ=3.0; (g) σ=3.5; (h) σ=4.0; (i) σ=4.5
Fig. 3. Quaternion Fourier transform spectra of Monarch blurred image under different standard deviations. (a) σ=0.5; (b) σ=1.0; (c) σ=1.5; (d) σ=2.0; (e) σ=2.5; (f) σ=3.0; (g) σ=3.5; (h) σ=4.0; (i) σ=4.5; (j) σ=5.0
Fig. 4. Principle of color image quality assessment
Fig. 5. Building blurred images and their corresponding spectra. (a)(c) σ=0.5; (b)(d) σ=5.0
Fig. 6. Original color images. (a) Bike; (b) building; (c) cap; (d) monarch butterfly; (e) painted house; (f) parrot
Fig. 7. Noiseless blurred images. (a)-(c) σ=5.0; (d)-(f) ρ=20
Fig. 8. Quality assessment results of noiseless Gaussian blurred images
Fig. 9. Quality assessment results of noiseless motion blurred images
Fig. 10. Noisy blurred images. (a)-(c) v=0.02; (d)-(f) d=0.20
Fig. 11. Quality assessment results of images with Gaussian white noises (v=0.01)
Fig. 12. Quality assessment results of images with Gaussian white noises (v=0.02)
Fig. 13. Quality assessment results of images with salt and pepper noises (d=0.10)
Fig. 14. Quality assessment results of images with salt and pepper noises (d=0.20)
σ | Tthres | TH | Sscore |
---|
0.5 | 132.3994 | 3783 | 0.0096 | 5.0 | 130.9453 | 738 | 0.0019 |
|
Table 1. Assessment results
Database | Parameter | CPBD | ENIQA | GSVD | JNBM | LPC-SI | QFTM |
---|
| PLCC | 0.9519 | 0.4843 | 0.7945 | 0.9368 | 0.9788 | 0.9230 | IVC database | SRCC | 0.8663 | 0.2534 | 0.9366 | 0.8965 | 0.9622 | 0.9252 | | KRCC | 0.7097 | 0.1613 | 0.7904 | 0.7420 | 0.8710 | 0.7742 | | RMSE | 0.3498 | 0.9988 | 0.6933 | 0.3994 | 0.2336 | 0.4393 | | PLCC | 0.8958 | 0.8558 | 0.8169 | 0.9142 | 0.9229 | 0.9480 | TID2013 database | SRCC | 0.9050 | 0.8618 | 0.8358 | 0.9213 | 0.9264 | 0.9437 | | KRCC | 0.7221 | 0.6874 | 0.6309 | 0.7566 | 0.7666 | 0.7966 | | RMSE | 0.5546 | 0.6455 | 0.7197 | 0.5056 | 0.4805 | 0.3973 | | PLCC | 0.8933 | 0.8896 | 0.8762 | 0.8888 | 0.8941 | 0.9463 | CSIQ database | SRCC | 0.9426 | 0.8892 | 0.8766 | 0.8603 | 0.9539 | 0.9514 | | KRCC | 0.8115 | 0.7051 | 0.6894 | 0.7237 | 0.8322 | 0.8178 | | RMSE | 0.1288 | 0.1309 | 0.1381 | 0.1313 | 0.1284 | 0.0926 |
|
Table 2. Performance comparison among algorithms on IVC, TID2013 and CSIQ databases
Method | CPBD | ENIQA | GSVD | JNBM | LPC-SI | QFTM |
---|
Running time /s | 0.1404 | 13.1112 | 0.4676 | 0.7123 | 0.7956 | 0.0791 |
|
Table 3. Average running time of methods