Wavelet Image Interpolation Using Approximate Modeling of Exponential Decay

Image interpolation in the wavelet domain is the estimation problem of the highest frequency subbands. The exponential function can approximate the decay of wavelet coefficients across scales. If we can determine a proper decaying model, the magnitudes of detail coefficients will be predictable. In this paper, we exploit a simple exponential model with meta-parameters which reflects the self-similarity of natural images. The experimental results show that the proposed method outperforms the previous wavelet-domain interpolation methods as well as the conventional bicubic method