1 of 11

  • I think the problem has been resolved
  • I have been doing the following analysis
    • darks � ↓�flatten (convert to 1d)� ↓�standardization (convert each pixel with 0.0 mean and 1.0 std)� ↓�PCA
  • Standardization is not necessary for 2d-images�Only necessary when each features have different properties

2 of 11

{20250123: [119780, 119782, 119785, 119787, 119789, 119791, 119795, 119799],

20250125: [120096, 120098, 120100, 120103, 120106, 120109, 120112],

20250127: [120579],

20250128: [120720, 120722, 120724, 120726, 120728]}

Frames has been split into 2 sub-samples to check the consistency of PCA results with each other after removing frames with undetected CRs.

{20250123: [119781, 119783, 119786, 119788, 119790, 119792, 119796],

20250125: [120094, 120097, 120099, 120102, 120105, 120108, 120110, 120113],

20250127: [120581],

20250128: [120721, 120723, 120725, 120727]}

3 of 11

PCA components for 1st group

Only 2 components are significant

4 of 11

PCA components for 2nd group

Only 2 components are significant

5 of 11

Difference of PCA components for 1st and 2nd group

PCs are the same between two groups. Small difference for the mean.

6 of 11

Residuals from reconstructed images.

2 components are used.

​

1st images using 1st PCA components

7 of 11

Residuals from reconstructed images.

2 components are used.

​

1st images using 2nd PCA components

8 of 11

Residuals from reconstructed images.

2 components are used.

​

2nd images using 2nd PCA components

9 of 11

Residuals from reconstructed images.

2 components are used.

​

2nd images using 1st PCA components

10 of 11

Residuals from reconstructed images.

1 components are used.

​

1st images using 1st PCA components

​

Only 1 component is not sufficient to correct the variation

11 of 11

Residuals from reconstructed images.

3 components are used.

​

1st images using 1st PCA components

​

There is no gain using 3 components