A weighting algorithm based on the gravitational model for local stereo matching

Abstract

In the local stereo matching method, the weight coefficient is vital to the cost aggregation of a given support region. In this paper, a novel weighting algorithm based on the theory of gravity is proposed, which is appropriately called the weighting coefficient based on the gravitational model. First, regarding the concept of gravity, the pixels are considered particles that have their own properties in a given support region. Second, the properties of pixels are defined by being combined with the color similarity, geometric proximity and matching cost in a given support region, in which the mass of each pixel is computed based on the matching cost, while the distance is calculated according to the color similarity and geometric proximity. Ultimately, a gravitational force model is developed among the pixels in a support region. In the presence of gravity, weight coefficients are assigned to each pixel. Accordingly, more reasonable weight coefficients are assigned to the pixels with incorrect corresponding pixels and the outliers pixels in a given support region, and a more accurate aggregated cost of the given support region is computed in the process of stereo matching. The experimental results show that in comparison with other available methods, the proposed algorithm can achieve more accurate disparity in various matching regions.