Topics in Unconventional Imagery

摘要

A new approach for digitally reducing the presence of random atmospheric turbulence in imagery was developed. This image division' method is based upon the use of two short-exposure images as data. These have unknown point spread functions (PSFs) as defined by the random turbulence. The latter are found by dividing the two image spectra (thus image division'), producing a set of linear equations which may be inverted for the PSFs. The inversion problem is inherently ill-posed. However, the solution is stabilized by imposing prior knowledge about the object and two PSFs that: (a) they have finite support extensions, and (b) they numerically obey positivity. Once the two PSFs are found, they are used to inverse-filter their corresponding images. The result is two output reconstructions of the object, which are simply averaged to produce the final output. The approach was successfully applied to both simulations and to real infrared imagery from Kitt Peak National Observatory.