Every image has noise. This noise is unavoidable, and greatly complicates the deconvolution process.
To provide the optimum deconvolution, MaxIm DL needs to know something about the noise and background level in the image. When these parameters are set properly, the algorithm converges more strongly towards the correct solution.
Two predominant noise sources are usually encountered. The first is photon shot noise – random variations in light itself. The Poisson noise model approximates this. MaxIm DL automatically calculates the noise level in each pixel based upon the number of photoelectrons detected in that pixel. In order to do this, it needs to know the gain of the camera, a value known as Photoelectrons per ADU or simply Gain (this can be determined using the Photons Wizard). The estimated amount of noise in a pixel simply corresponds to the square root of the number of photons detected.
Another noise source is the electronic read noise. On average, the same amount is present in every pixel, so it is usually modeled using a Uniform noise model. To obtain a measure of the uniform noise level (i.e. the standard deviation of the distribution), the variance in pixel values can be measured in some part of the image that contains no foreground data. Often it is best to average the results from several locations. If the image has no background, due to high light levels, then you may need to measure the noise in a dark frame.
Deconvolution also requires an input parameter that indicates the average background level in the image. Even for astronomical images, the background is almost never black (zero) due to sky glow, light pollution, and other effects. Sometimes an image has had a background level previously added or subtracted; in that case, this amount (the black level offset) should be supplied so that the algorithm knows where the true zero point is. MaxIm DL provides easy-to-use tools to extract the necessary information from the image itself.