Image Interpolation Based on Statistical Relationship Between Wavelet Subbands

Wavelet-based image interpolation is a process of estimating the finest detail coefficients from the coarser scale coefficients. Wavelet coefficients can be regarded as the realization of a given distribution. Hence, if we can estimate the probability density function (pdf) of the coefficient from that of the coarser scale coefficient, the estimation of the finest detail coefficients will be also possible. In this paper, we propose the technique of estimating the parameters of the coefficient pdf for the purpose of image magnification. Parameter estimation will be described based on the approximation of wavelet statistics and properties. Simulation results show that the proposed scheme outperforms the previous wavelet subband extrapolation methods based on the estimation of the probability model.