In this paper, we propose a new image interpolation method based on a 2-D piecewise stationary autoregressive (PAR) model. Soft-decision adaptive interpolation (SAI) and its improved methods which are based on PAR model are state-of-the-art methods in image interpolation. They produce upconverted images with good quality but tend to have high calculation cost because of solving the inverse matrix problem to estimate the weighting parameters. So we use Gaussian function to estimate parameters without the inverse matrix calculation and reduce the calculation cost. Moreover, parameters are estimated on pixel by pixel, while in SAI they are constant in a local window. Thanks to these, the proposed method outperforms SAI in the quality of upconversion with low calculation cost. Finally we validate the proposed method through some experimental results.
展开▼
