An Improved Region-Growth Algorithm for Disparity Estimation

摘要

In this paper, we present a simple and efficient disparity estimation method based on region-growth techniques, which can obtain the more accurate disparity map from two color or gray images with an acceptable speed and computationnal cost. The proposed method includes two steps: the selection of the seed points and propagation. Instead of propagating after getting all seed points, the proposed method begins to propagate as soon as one point is gotten. This makes the algorithm more efficient. In the propagation step, a line-growth method is adopted since disparity of the stereo image is only in the row directions. Experiments with test and real stereo image pairs show the validity of the proposed method.