一种基于光流和能量的图像匹配算法

Images Matching Approach Based on the Optical Flow and Energy

结合光流与图像信息,提出一种获取稠密视差的图像匹配算法。首先对于基线较大的左右图像,在多分辨率框架下采用由粗到精的策略计算光流,从而实现大偏移量时的光流获取。其次为了避免光流在图像边界上的不可靠性,通过光流计算所得的光流场作为初始视差图,采用基于能量的方法依据对应的图像梯度场对光流场内部进行平滑并保持边缘的不连续性,最终得到精准稠密的视差图。实验验证,该方法是一种行之有效的图像匹配算法。

An image matching algorithm combining optical flow with image information is proposed in this paper to acquire dense disparity map. Firstly, in the multi resolution frame the coarse-to-fine strategy is applied to calculate the optical flow for wide baseline stereo pairs, so the large displacements optical flow estimation is realized. Secondly, in order to solve the unreliability of the optical flow at the image edge, taking the optical flow field as the initial disparity map,energy based approach is adopted that smoothes disparity inside the boundaries and remains the discontinuity across the boundaries according to the gradient field of image, then dense and precise disparity map is acquired. Finally, the pro- posed image matching algorithm is illustrated with three stereo image pairs, the validity and feasibility of our approach is experimentally verified.