Motion Deblurring Using the Similarity of the Multiscales

Haisen Li,Yanning Zhang,Jinqiu Sun

DOI: https://doi.org/10.1016/j.ijleo.2014.09.014

IF: 3.1

2015-01-01

Optik

Abstract:Motion blur exists widely when the image is captured using a hand-held camera, especially when the exposure time is long. Many motion deblurring methods have been proposed to estimate a satisfying latent image; however, these always exist some artifacts in the deblurring result. In this paper, via introducing the similarity of multiscales, a blind image deblurring method is proposed. Considering similarity of the latent image between different scales, it is used as a prior to constrain the estimated latent image similar as pre-scale to reduce artifacts which result from deconvolution. Experiments indicate that the proposed method could obtain better performance than the state-of-the-art methods, and could obtain satisfying deblurring result with fewer artifacts.