A stereo matching algorithm based on fuzzy identification

Stereo vision has long been one of the central research problems in computer vision, and stereo matching is the most important and difficult issue of stereovision. We propose an algorithm by fuzzy identification in fuzzy theory to cope with the uncertain and fuzzy characteristic during matching. For image pairs with dense distinct features, we adopt a feature-constrained strategy. First, we obtain a precise disparity map of edges using fuzzy logic. Next we fill in the disparity data of other points under the constraint of the acquired disparity map. For images with sparse features, we use intensity segments as elements to be matched and introduce the principle of DTW, which is applied for speech recognition, into the matching processes on the segments. Experiments with real and synthetic images have been performed to demonstrate the effectiveness of the algorithms.