Information Technology Reference
In-Depth Information
Table 4
RMSE of IQMs on stereoscopic images
IQM
Blurring
JPEG
JPEG2000
Noise
All
PSNR
1.97
5.97
5.09
2.74
7.64
SSIM
3.00
7.70
8.91
2.43
9.28
MSSIM
1.91
4.94
4.97
2.51
7.62
VSNR
2.58
5.71
5.42
3.54
8.63
VIF
1.84
4.39
6.45
3.41
7.78
UQI
1.89
3.96
6.44
4.06
7.13
IFC
1.78
3.55
6.37
3.98
8.61
NQM
2.17
3.53
4.23
4.36
8.70
WSNR
2.02
6.74
6.09
3.85
9.07
PHVS
2.10
5.62
5.27
2.65
8.06
JND
2.18
6.97
5.73
3.55
8.58
Average
2.13
5.38
5.91
3.37
8.29
Table 5
Pearson correlation coefficients of IQMs on stereoscopic images
IQM
Blurring
JPEG
JPEG2000
Noise
All
PSNR
0.939
0.882
0.950
0.978
0.795
SSIM
0.851
0.793
0.824
0.983
0.677
MSSIM
0.943
0.920
0.948
0.981
0.797
VSNR
0.893
0.893
0.948
0.962
0.731
VIF
0.947
0.938
0.916
0.966
0.788
UQI
0.943
0.950
0.913
0.950
0.825
IFC
0.951
0.962
0.929
0.954
0.734
NQM
0.925
0.961
0.963
0.942
0.726
WSNR
0.935
0.846
0.924
0.955
0.696
PHVS
0.930
0.896
0.949
0.979
0.769
JND
0.924
0.836
0.937
0.961
0.738
Average
0.926
0.898
0.927
0.965
0.752
Subsequently, we performed the IQMs on the right eye images in the stereo-
scopic image quality assessment, as there were no distortions on the left eye
images. Table 4 and Table 5 give the evaluation results. In addition, we can use a
constant as the quality value on the left eye images, e.g. 1, for SSIM, MSSIM, UQI,
etc. Then, the significance of the interaction effects between the quality values of
the left eye images, the right eye images, and the overall qualities was tested by
performing a two-way ANOVA (ANalysis Of Variance) on the results. The results
of the ANOVA show that the quality of the right eye image dominates the overall
quality and there is almost no influence of the left eye image on the overall quality.
In addition, according to the evaluation results, the situation of the performance of
