Novel Demosaicking Method Using Nonlocal Similarity Fusion

Although most demosaicking methods assume the existence of high local correlation in estimating the missing color components, such an assumption may fail for images with high color saturation and sharp color transitions. This paper presents a demosaicking scheme by exploiting both the variance of color differences (VCD) and the non-local similarity. First, the missing green components are estimated according to VCD along different edge directions. Then, the nonlocal pixels similar to the estimated pixel are searched to improve the initial estimate of the G channel. Based on the interpolated green plane, the missing blue and red components are preliminarily estimated. Finally, the blue and red channels are enhanced by exploiting nonlocal redundancies respectively. Experimental results show that the proposed algorithm is able to improve the CPSNR, sharpen edge and texture and lead to higher visual quality of reconstructed color images.