CLEAN (algorithm)

The CLEAN algorithm is an iterative deconvolution method for images made in radio astronomy, published by Jan Hogbom in 1974. A radio telescope array does not record the sky directly. What it records, called the dirty image, is the true sky blurred by the instrument's point spread function, known as the dirty beam. CLEAN assumes that the sky is made of a number of point sources. It repeatedly finds the brightest value in the dirty image, subtracts a small fraction of that peak, scaled by a gain, convolved with the dirty beam, and records the removed amount as a component of the true image. The loop stops when the highest remaining value falls below a chosen threshold. The astronomer T. J. Cornwell wrote that the impact of CLEAN on radio astronomy has been immense, both by speeding up and sharpening observations and by encouraging a wave of innovation in synthesis processing. Several variations have been proposed since 1974, and the method has also been applied in other areas of astronomy and science. Newer work such as the Event Horizon Telescope imaging of the M87 black hole used regularized maximum likelihood, which estimates an image forward from an objective function instead of deconvolving.