Norbert Wiener published Extrapolation, Interpolation, and Smoothing of Stationary Time Series in 1949. This book laid the groundwork for deconvolution theory at the Massachusetts Institute of Technology.
The lower the signal-to-noise ratio, the worse the final estimate becomes. If a noisy image assumes no noise exists, statistical estimates become incorrect and amplify errors.
Enders Robinson worked with Norbert Wiener and Norman Levinson at MIT in 1950 to develop the convolutional model for reflection seismograms used to map Earth structure.
Blind deconvolution deduces this function by systematically testing different possibilities until image quality improves. Finding the true point spread function remains impossible, so approximations based on known probes are used instead.
The CLEAN algorithm serves as a commonly used method for this specific type of astronomy work involving deconvolving the produced image with a dirty beam.