Inverse Halftoning via Median Interpolating Pyramid

Work along with the property of error diffuse halftoning a multi-scale interpolating pyramid transform on two dimension is constructed with the nonlinear median operator. Having investigated the transform distribution of halftone noise in the approximation and the detail on each-scale, the de-noising method is studied through many experiments, and an inverse halftoning algorithm based on the expanded median-interpolating pyramid transform (EMIPT) is proposed. Firstly the method decomposes halftoning image by the transform in one-scale, and a Wiener filter is applied to the detail coefficient. Then an inverse image is reconstructed from the cleaned detail and the approximation. Experiments show that the proposed transform can separate the halftone noise and the image detail better than linear multi-scale transform. And the inverse halftoning method based on it includes some excellent performances such as smoothing the halftone patterns, efficient edge preserving ability, high PSNR and small memory requirement