Problems
- Dark current - without calibration, the zero point of the image isn’t correct
- Pixels are captured by different photodiodes → they don’t have the same properties
- Vignetting effect
- Dust particles
Definitions
Input image - A(t, T, g)
- t - exposure time
- T - chip temperature
- g - gain (not relevant for CCD)
Dark Frame - D(t, T, g) Flat-field image - F(t, T, g) Calibrated image -
- Upper part of the fraction - deletes the |dark current and bias influence
- Lower part - takes care of Vignetting effect, non-same sensitivity and dust particles
- k - constant based on image bit resolution
- f - offset
Calibration by itself causes the noise level to go up. For that reason you use Master Dark Frame and Master Flat-fields. Both are created by combining multiple Dark Frames or Flat-fields. How to create them:
Mean - for one image, for n images it is
Creates impulse noise - bad pixels on random positions, mainly caused by cosmic radiation. The number of bad pixels increases with altitude.