Spatially adaptive video restoration using truncated constrained least-squared filter

Abstract

In this paper, an adaptive video restoration method is presented for removing spatially-varying blur using truncated constrained least-squared (TCLS) filter. The proposed method consists of two modules: i) spatially-varying blur estimation based on blur map optimization in temporally adjacent frames and ii) adaptive image restoration using TCLS filter according to the estimated blur parameters. The proposed method can restore a video without artifacts by estimating the optimal spatially varying blur map, and the use of the TCLS restoration filter enables fast video restoration. Experimental results show that the proposed method can better restore the test video by 1.2 times in the sense of peak-to-peak signal-to-noise ratio (PSNR).