Noise reduction scheme of temporal and spatial filter for 3D video real-time processing

In this paper, we suggest a motion adaptive noise elimination method in moving pictures, which combines temporal filtering and spatial filtering for real time processing of 3D images. To realize 3D pictures, two sets of images taken by a two-lens camera are rectified. And then, a disparity map is generated using binocular parallax of the rectified two sets of images. At this point, temporal filtering and spatial filtering are applied to generate a disparity map with less noise that also enables real time processing of obtaining high quality 3D images. Temporal filtering corrects errors occurred in the disparity map. Spatial filtering clarifies the boundaries of objects from the background, calibrates images again to reduce noises or surrounding error, and reinforces the edges lost in the images, to get a disparity map of high performance. Our approach showed 6.4dB increment in image quality, realizing the algorithm using RTL(Resister Transfer Level) in real time processing.