Statistical Filtering of Images in the Spectral Domain

The problem of increasing the efficiency of search and identification of anomalies in multispectral images obtained in observation systems is investigated. The approach based on the use of the spectral form of the mathematical description of the observation system model and statistical filtering of video information performed in the spatial-frequency domain is considered. The method is based on finding the spectral characteristic of the image and filtering operators, transition to solving the system of linear algebraic equations of generalized characteristic functions and estimation of the posterior distribution density of the image state vector. To perform optimal statistical filtering and image anomaly detection, an algorithm of recurrent processing of video information by a multichannel frequency filter is implemented, providing maximum suppression of noise additively distorting the useful image signal #CSOC1120.