Statistical characteristics of discrete cosine transform (DCT) coefficients in blocks of grayscale test images rae analyzed. Robust estimators of scale and tail heaviness are used in analysis since it is shown that the distributions are non-Gaussian and they are better described by generalized Gaussian distribution with variable parameter. The parameters of these distributions are determined for noise-free images and images corrupted by additive white Gaussian noise. It is shown that the presence of noise considerably changes the distributions that approach to Gaussian for high-frequency DCT coefficients. Parameters of the generalized Gaussian distribution depend on both image properties and noise variance.
DCT-coefficients, image blocks, statistical characteristics, generalized Gaussian distribution