Image interpolation algorithm based on learning edge information
Wang Bei-bei · Jisuanji gongcheng yu sheji · 2009
In order to solve borderline blurred and double shadow phenomenon produced by image interpolation, gray-level image interpolation is realized using neural network learning image edge information, and matching similar pixel between adjacent cross-image.For each pixel on new interpolating middle image, first a series of candidate pixel corresponding are selected on two adjacent cross-image, then the best a brace of pixel from between candidate pixel group is matched by neural network, at last interpolated into a new gray scale pixel value on corresponding position on new interpolating middle image by using this a brace of pixel.Experimental results show that the algorithm improve borderline blurred phenomenon, and also eliminate double-shadow phenomenon greatly produced by traditional image interpolation.