A study on the relationship between depth map quality and stereoscopic image quality using upsampled depth maps - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 128

Table 2
Average Objective Measurement Data of Upsampled Depth Maps (PSNR unit: dB)

Depth Map BLU BCU BU JBU VBU ABU DTBU

PSNR 35.85 35.71 35.64 34.15 35.64 33.16 34.86

E-PSNR 23.68 23.55 23.66 22.82 23.38 20.97 22.93

NE-PSNR 38.07 37.94 37.78 37.50 37.93 35.43 36.92

Sharpness 39.5 42.2 49.51 49.09 31.92 88.31 68.14

Blur 8.48 11.38 10.29 10.87 10.51 9.00 9.89

SSIM 0.976 0.955 0.975 0.956 0.971 0.962 0.972

VIF 0.518 0.539 0.424 0.422 0.478 0.398 0.438

BIQI 57.8 66.34 63.11 32.81 41.94 29.15 72

NIQE 15.95 13.11 13.94 11.82 12.47 13.41 13.82

The 3D perception grades of upsampling methods in Table 1 are based on 3D visual discom-
fort.
Quality scores of upsampled depth maps obtained from each IQA metric are considered
as a group of seven samples. All values are normalized by scaling between 0 and 1 and the
similarity of samples distribution in each IQA group is compared with subjective evaluation
samples group using Pearson, Spearman, and Kendall correlation coefficients. Table 3 shows
the correlation results.

Table 3
Pearson, Spearman, and Kendall Correlation Coefficients Between Subjective and Objective
Measurements

NE-
PSNR

Blur
Metric

PSNR E-PSNR Sharpness SSIM VIF BIQI NIQE

Pearson 0.528 0.608 0.554 − 0.522 0.273 0.505 0.019 − 0.34 0.132

Spearman 0.035 0.142 0.035 − 0.321 0.142 0.357 0.107 − 0.321 0.107

Kendall 0.047 0.142 0.047 − 0.142 0.047 0.142 0.142 − 0.238 0.142

Before evaluating the strength of correlation using different correlation coefficients, it is
worth mentioning that Pearson's correlation coefficient takes into account both the number
and degree of concordances and discordances, whereas Kendall's tau correlation coefficient
shows only the number of concordances and discordances. Spearman's correlation is in
between of the Pearson's and Kendall's, reflecting the degree of concordances and discord-
ances on the rank scale. The disadvantage of Pearson is the sensitivity to outliers (an observa-
tion that is numerically distant from the rest of the data). In this case, Spearman and Kendall
are less sensitive to outliers and preferable.