USING WEIGHTED MEDIAN FILTERING FOR FAST IMAGE SUPER-RESOLUTION Nasonov A.V., Krylov A.S. Lomonosov Moscow State University, Faculty of Computational Mathematics and Cybernetics, Laboratory of Mathematical Methods of Image Processing http://imaging.cs.msu.ru A method of image super-resolution (SR) is presented -- reconstruction of a high resolution image from several low resolution images. A non-iterative method based on weighted median filtering is proposed. The super-resolution is posed as a system of equations Ak z uk , k 1,2,..., N , (1) where z is the high-resolution image, u k are the given low-resolution images, Ak z DFk Hz is a downsampling operator [1], where D is the decimation operator, H is the Gauss filter, Fk is the motion operator. We assume, that the motion is known. In the proposed algorithm, the problem (1) is formulated in the following form: ( Hz)( xn , yn ) wn and consists in reconstruction of the blurred high-resolution image Hz with known values wn in the given points ( xn , yn ) . Even small errors in motion estimation results in serious degradation of the reconstructed image. To make the method stable to errors in the motion vectors, an averagi ng is used. For every target pixel (i, j ) , several values wn from a certain neighborhood of (i, j ) are taken. The following averaging methods are considered: Gauss filter [2], median filter [3] and the proposed weighted median filter which is a combination of Gauss filter and median filter. To estimate to results, edge adaptive metrics (BEP and BEN) from [4] were used. Illustrations of super-resolution methods for scale factor 2 and 16 low-resolution images are given. It is shown that the use of weighted median averaging enables us to reduce the influence of errors in motion vectors calculation. The work was supported by federal target program Scientific and scientific-pedagogical personnel of innovative Russia in 2009-2013 and RFBR grant 09-01-92474-MHKC. 1. 2. 3. 4. Literature S. Farsiu, D. Robinson, M. Elad, P. Milanfar Fast and Robust Multi-Frame SuperResolution // IEEE Trans. Image Processing, Vol. 13, No. 10, 2004, pp. 13271344. A.S. Krylov, A.V. Nasonov, O.S. Ushmaev Video super-resolution with fast deconvolution // Pattern Recognition and Image Analysis, 2009, Vol. 19, No. 3, pp. 497-500. A. SАnchez-Beato, G. Pajares Robust Super-Resolution Using a Median Filter for Irregular Samples // Lecture Notes in Computer Science, Vol. 5524, 2009, pp. 298305. A.V. Nasonov, A.S. Krylov Adaptive Image Deringing // Proceedings of GraphiCon'2009, Moscow, Russia, October 2009, pp. 151-154.