Gaussian DCT coefficient models (Q1027805)

It is assumed that the discrete cosine transform (DCT) coefficients of most natural images follow a Gaussian distribution. In this theoretical work, the author obtains sixteen compounded distribution models from \[ f(x)=\frac{1}{\sqrt {2\pi}}\int^\infty_0\frac{1}{\sigma}\exp\left(-\frac{y^2}{2\sigma^2}\right)g (\sigma^2)d\sigma^2 \] by taking different density functions \(g(\sigma^2)\) of \(\sigma^2\). The parameters, in each case, are estimated by the method of moments and the maximum likelihood method. It is shown, by using Box plots, that the new models are better than the Gaussian model for image analysis and image coding.