Table 1 Selected summary statistics from the PSNR gain distributions shown in Main Fig. 2c

From: Robust self-supervised denoising of voltage imaging data using CellMincer

Algorithm

Q

Mean PSNR

Mode PSNR

IQR

CellMincer

5

23.25

23.66

0.75

10

23.02

23.67

0.80

50

22.76

23.75

1.41

200

20.50

22.12

2.02

SUPPORT

5

23.10

23.16

1.57

10

22.29

22.66

1.80

50

20.71

20.82

2.12

200

19.21

19.80

2.42

PMD

5

22.01

22.16

1.14

10

21.64

21.88

1.23

50

20.58

21.11

1.50

200

19.43

20.76

1.68

DeepCAD-RT

5

16.41

16.86

0.79

10

15.01

15.40

0.70

50

12.68

13.37

0.59

200

10.59

11.67

0.56

N2S

5

8.61

8.38

0.64

10

8.47

8.61

0.78

50

8.26

7.99

0.63

200

7.02

6.24

1.40

DeepSeMi

5

9.80

9.74

0.37

10

8.76

8.86

0.40

50

11.81

13.75

1.70

200

9.31

12.71

2.90

DeepVID

5

−1.69

−1.33

0.72

10

3.33

3.75

1.08

50

−15.45

−15.13

0.58

200

−10.37

−10.00

0.53

  1. Both the distribution mean and mode are included to contrast the central PSNR tendency, which is influenced by outlier frames, with the expected denoising outcome for a “typical” frame. The interquartile range (IQR), included as a representative statistic for distribution spread, was calculated using the distribution mode as its center.