Color demosaicking by local directional interpolation and nonlocal adaptive thresholding

摘要

Single sensor digital color cameras capture only one ofnthe three primary colors at each pixel and a process called colorndemosaicking (CDM) is used to reconstruct the full color images.nMost CDM algorithms assume the existence of high local spectralnredundancy in estimating the missing color samples. However, fornimages with sharp color transitions and high color saturation, suchnan assumption may be invalid and visually unpleasant CDM errorsnwill occur. In this paper, we exploit the image nonlocal redundancy tonimprove the local color reproduction result. First, multiple local direc-ntional estimates of a missing color sample are computed and fusednaccording to local gradients. Then, nonlocal pixels similar to the esti-nmated pixel are searched to enhance the local estimate. An adaptiventhresholding method rather than the commonly used nonlocal meansnfiltering is proposed to improve the local estimate. This allows thenfinal reconstruction to be performed at the structural level as op-nposed to the pixel level. Experimental results demonstrate that thenproposed local directional interpolation and nonlocal adaptive thresh-nolding method outperforms many state-of-the-art CDM methods innreconstructing the edges and reducing color interpolation artifacts,nleading to higher visual quality of reproduced color images