3D image denoising using stereo correspondences

In this paper, we propose an image denoising algorithm for 3D demosaicked images. Each color component of the 3D stereo images is processed for noise reduction by identifying the stereo correspondences between two stereo images at subpixel accuracy. Using our proposed method, it has been experimentally shown that there is visual improvement in the demosaicked output images using a realistic signal-dependent noise model with demonstration of a higher signal-to-noise ratio (SNR).