Adaptive Filtration of Multispectral Images

Abstract

The algorithm of adaptive smoothing of images subject to their gradients is offered. It is based on accurate calculation and automatic binarization of the gradient and adaptive Gaussian smoothing of images with preservation of borders of homogeneous regions. The algorithm is developed specially to process noisy aeroand the space images that usually contain a large number of objects of small size. Local characteristics of the noise and the gradient are used to determine adaptive window for each pixel. The windows of adaptive smoothing are fitted so that they do not cross binarized gradient.