A method for simultaneous outlier rejection in image super-resolution

Abstract

In this paper, we propose a method to adaptively reject outlier image regions in the process of super-resolution image reconstruction. We use adaptive FIR filtering while iteratively fusing the gradient images. The LMS adapted filter coefficients automatically isolate the outlier image regions, for which motion was inaccurately estimated. The adaptation criterion used is the median of the errors at each pixel location. Through simulated experiments on synthetic images, we show that the proposed technique performs well in the presence of outlier images. This relatively simple and fast mechanism enables to add robustness in practical implementations of super-resolution, while still effective against Gaussian noise. © Springer-Verlag Berlin Heidelberg 2003.