A saliency based noise reduction method for Digital TV

Abstract

In this paper, we propose a content-aware noise reduction (NR) method for the Digital TV, which preserves image textures and edges. In the proposed method, noise is first reduced by a temporal recursive NR filter and then further suppressed by a spatial NR filter. To preserve the image texture, the saliency map which accurately represents edges in a noisy image is employed in the spatial filter. Experimental results show that the proposed method exhibits the better NR performance as compared with conventional methods.