Robust noise reduction using a hierarchical motion compensation in noisy image sequences

Abstract

In this paper, robust noise reduction method is proposed to stably reduce the noise by applying the adaptive temporal IIR filter using the hierarchical motion compensation. Motion vector of each pixel is obtained by hierarchical motion estimation which is reducing template block size from 8×8, to 4×4 and to 2×2. With obtained motion vector, filter weight is adjusted adaptively according to the correlation between the related pixels which is determined considering not only pixel similarity to temporal axis but also the neighboring pixel variation in spatial domain.