Data Reduction with Skynet

Part 1

Open and inspect the files in the following subfolders within one of your User Observations folders:

  • Raw Images: This is the preprocessed image, fresh off the telescope with no calibrations applied
  • Master Calibration Images: These are calibration images taken to remove sources of noise in the raw image. There are three calibration images,  corresponding to three different corrections made to the data:
    • bias image (or frame) corrects for readout noise, from the read-out amplifier electronics
    • dark frame corrects for thermal electrons in the image, i.e. counts that don’t come from photons but from thermal electrons in the conduction band
    • flat field corrects for differences in sensitivity across the chip, including obscuration by dust particles which look like big donuts because they’re out of focus!
  • Reduced Images: This is the final version of the image, processed for you by the Skynet algorithms, using the calibration images

What is the typical pixel count  in the bias frame? 

In the dark frame?

In the flat field?

What is the exposure time of the bias frame? 

The dark frame?

The flat field?

What is the exposure time of the raw image frame?

Part 2

How to correct for the bias, dark, and flat using image math.

Image math refers to the fact that we can perform operations on the digital images on a pixel-by-pixel basis. The two basic types of operation we’ll work with are:

  1. Scalar:  A scalar operation is applied to each pixel in an image, e.g. every pixel value is divided by 10 in the resulting image
  2. Image: Two images are involved in the operation, pixel-by-pixel, e.g. Image1 + Image2 means that the value of pixel (1,1) in the resulting image is the value of pixel (1,1) of Image1 plus the value of pixel (1,1) in Image2 (same for pixel (1,2), pixel (1,3) and all other pixels)

For our image reduction we want:

Final image = (Raw_image – bias_frame – (dark_frame)(exposure_time_scale) ) / (normalized_flat_field_frame)


  • exposure_time_scale is a scalar = (exposure time of raw image)/(exposure time of dark current frame)
  • normalized_flat_field_frame is an image = flat_field_frame/(mode of flat field frame)
    recall that the mode of a set of numbers (pixel values here) is the most frequently occurring value

More information: A basic First-Aid procedure for CCD images