A Krawtchouk moments based super resolution technique for multiframe image sequence

Recently, Xinbo Gau et. al[2] employed Zernike moments to construct a high resolution image sequence from a given set of blurred, down sampled and noisy image sequence. In general, Zernike moments are defined using Zernike polynomials which are not orthogonal over the square image region. Further, these moments are not discrete. Hence, image quality gets distorted due to incorrect moment values. Therefore, to correct these errors, this paper uses Krawchouk moments that posses the following features i) no need for coordinate transformation ii) orthogonal over a square region and iii) they are discrete moments, to create a high resolution image sequence from a given low resolution image sequence. The proposed approach was tested using different multiframe image sequences having small local motion and rotation. Results are compared with Zernike moments, Lanczos and sparse representation methods.