The calibration images bias, dark, flat and darkflat-frames
In astrophotography, influences of the camera and the equipment often cause several overlapping disturbances in single exposures (sub-frames or also light-frames). Some can be eliminated by additional correction images. How this is mathematically solved for a sub-frame in the software is shown in the following illustration.
Schematic illustration of the use of the calibration images in the creation of a corrected light frame, using the Cirrus Nebula as an example
Bias-Frame
Bias-frames are primarily needed to remove the offset value from the images. Since the offset value is not part of the object signal and is fixed, this value can be subtracted from the actual signal. A bias-frame contains the offset value set by the manufacturer and the one you set previously (see section 'Basics' - 'Offset') as well as readout noise. Since readout noise is a random value for every frame type, it is not eliminated by subtraction. To prevent the total noise from increasing unnecessarily due to the different readout noise values resulting from the subtraction, as many bias-frames as possible are averaged.
A bias-frame is created using the camera's shortest possible exposure time, which is usually in the µs-range. With such a short exposure time, dark current is virtually nonexistent.
The offset value is included in the calibration images (flat- and dark-frames) as well as the light-frames. Which of these is the most suitable point at which to apply the bias-frames? It should be noted that the offset value is subtracted only once during calibration. However, when the dark-frames are subtracted from the light frame, the offset value contained therein is already automatically subtracted.
(light + offset) – (dark + offset) = light – dark
So if the offset value were subtracted from the dark-frame, it would no longer be possible to eliminate the offset value from the light-frames.
(light + offset) – (dark + offset – offset) = light + offset – dark
If the offset value were subtracted from the light-frames as well as the dark-frames, this would be an unnecessary mathematical operation because it would yield the same result as without this subtraction, with the drawback that, as described above, an averaged value of the readout noise from the bias frames remains, which then affects the light- and dark-frames.
(light + offset – offset) – (dark + offset – offset) = light – dark + averaged readout noise
The master-flat- frames are also unsuitable for this purpose because they are divided by the light-frame corrected by the master-dark-frame, rather than subtracted from it.
But why is the bias-frame still needed? Since each flat-frame also contains the offset value, the flat-frames must be corrected for this offset value before they are used to calibrate (by division) the individual image. Otherwise, the master-flat-frame and the light-frame would not have the same offset basis, and the calibration would fail. If the flat-frames have a longer exposure time (e.g., with narrowband filters), the bias-frames should be replaced by darkflat-frames (see below). Due to the longer exposure time, dark current increases, and with some cameras, initial amplifier glow also becomes apparent; both of these can then be eliminated by the darkflat-frames.
Procedure for capturing bias-frames:
- Set the shortest exposure time in the software program
- Put the cover on the telescope or camera - no light of any kind must reach the chip
- The images must have the same ISO/Gain setting as the light- and flat-frames.
- The chip temperature does not matter for these capturing, so it does not hurt if the chip is already cooled.
- Take 20 to 50 captures, which are then averaged.
After stretching (stretching the tone curve), the individual pixels display the typical range of gray levels, which consist solely of the offset value and readout noise. After stacking, the bias-frames contain the offset value along with the averaged readout noise.
Dark-Frame
Dark frames are created to remove dark current, sensor glow, and hot- and dead- pixels from the light- or sub-frame. Dark-frames are subtracted from the sub-frame. At the same time, the sub-frames are cleared of the offset value, since this is contained in the dark-frames.
Hot pixels are created by single electrons jumping from one pixel to the neighboring pixel and react disproportionately to the additional energy. As a result, they glow red, yellow or sometimes green or blue in the final image. (https://www.fotografiewissen.de/technik/allgemein/deadpixel-und-hotpixel/)
In the case of dead-pixels, there is a technical problem so that there is no or a permanent current flow. As a result, the pixels are either black or white (burned out). (https://www.fotografiewissen.de/technik/allgemein/deadpixel-und-hotpixel/)
The sensor glow occurs because nearby electronics emit heat radiation in the infrared range and the pixels next to them react to it. (https://de.wikipedia.org/wiki/Sensorgl%C3%BChen)
180 s dark frame (stretched (stretching the tone value curve) with the program PixInsight)
Procedure for capturing dark-frames:
- Put the cover on the telescope or camera - no light of any kind must reach the chip
- Dark frames must be taken with the same exposure time, chip temperature and ISO/Gain as the sub-frames.
- With uncooled cameras it is difficult to achieve the same chip temperature. Processing such dark frames should not be used with uncooled cameras with weak sensor glow, as they lead to poorer image results.
- Take 15 to 30 captures, which are then averaged.
If a cooled camera is used, a dark frame library with different exposure times and ISO/Gain can be created, which only needs to be updated from time to time. This eliminates the need for time-consuming imaging at each night of shooting.
Flat-Frame
Flat-frames are intended to correct vignetting and uneven image illumination caused by dust or dirt in the optical system. In addition, with even illumination during flat-frames, the processing software can detect how differently the pixels react to light.
Vignetting usually occurs when several components are present in a beam path and one of them crops the edge of the image. As a result, the side edges of the image are darker in contrast to the center.
Procedure for capturing Flat-Frames:
- The camera must be in the same position as when capturing the light- or sub-frames - avoid rotating the camera.
- The focus point should be almost identical to the focus of the single frame exposures.
- The same ISO/Gain value must be set as for the sub-frames or dark-frames.
- The chip temperature plays a subordinate role for short-exposure flat frames. However, with longer exposed flat-frames, i.e. when slow optics are present or narrow band filters are used, the dark current comes into play again. - Cooling is then an advantage. For such flat-frames it is therefore important that not the bias-frame, which does not contain any dark current, is subtracted, but a darkflat-frame.
- A flatfield mask is recommended for uniform illumination. This is positioned tilt-free in front of the telescope aperture.
- The exposure time should be selected in such a way that the mountain of the histogram curve is centered at 50 % in the imaging program. - This means that the pixels all have a medium saturation.
- Some astro cameras have difficulties to expose only 1 s at a bright homogeneous illumination. Here it is helpful to increase the exposure time (e.g. to 3 s) and to reduce the illumination by the flatfield mask, so that the histogram curve is again at 50 %.
- Take 15 to 30 captures, which are then averaged.
Darkflat-Frame (auch Flatdark-Frame)
Darkflat-frames (also called Flatdark-frames) are used to remove the offset and the dark current from a longer exposed flat-frame. In general, it is always a good idea to use this type of calibration instead of bias-frames, because even short-exposure flat-frames of approx. 1 s can have a certain dark current under certain circumstances. It often happens that CMOS cameras output unsuitable signal values with very short exposed flat-frames, so that it is always appropriate to create longer exposed flat frames (> 1 s) and thus also to use darkflat-frames.
Procedure for capturing Flat-Frames:
- Exactly the same exposure time is used as for the flat-frames.
- The ISO/Gain value must also match the sub- or flat-frame.
- The chip temperature plays a subordinate role as with the flat-frames, but cooling is again very advantageous here.
- Instead of the flat field mask, the telescope is closed again as with the bias- and dark-frames.
- Take 15 to 30 captures, which are then averaged.