Table 1 Performance statistics of displacement estimation methods in simulated data

From: Estimating full-field displacement in biological images using deep learning

Method

PSNR (dB)

SSIM (/100)

LPIPS (/100)

Rewarp (/100)

Time (s)

DIC

24.6 (0.2)

68.6 (0.7)

92.6 (0.4)

97.1 (0.3)

128.6 (0.3)

OF

19.9 (0.3)

34.8 (1.4)

87.1 (0.7)

95.6 (0.5)

2.62 (0.01)

FlowNet

18.3 (0.2)

14.3 (0.8)

82.2 (0.8)

86.5 (1.0)

0.028 (0.0007)

StrainNet

19.5 (0.3)

55.5 (1.7)

89.0 (0.7)

92.8 (0.8)

0.389 (0.0006)

DEFORM-Net

26.5 (0.2)

73.7 (0.5)

96.4 (0.2)

98.0 (0.2)

0.068 (0.0008)

  1. Data is presented as mean value and standard error of mean in brackets (n = 50). Highest performance values within standard error bounds are highlighted in bold. PSNR peak signal-to-noise ratio of MSE; SSIM structured similarity index metric; LPIPS learned perceptual image patch similarity; Rewarp: Pearson correlation coefficient of rewarp loss.
Back to article page