Spherical Wiener filter

A novel group-theoretic method is presented for denoising a three-dimensional scene in isotropic noise. Images of the scene at varying depths are regarded as reference stochastic processes on the unit sphere to formulate Weiner-Hopf equations for estimating the image at any given depth. These comprise a set of coupled linear integral equations on the unit sphere and are solved using Peter-Weyl theory of Fourier transform on the rotation group. The computational complexity of this algorithm is reduced using bi-invariance of the image correlations with respect to the stabilizer subgroup of the rotation group.