【摘要】 提出了两种方法来提高图像匹配的精度。一种方法是改变自适应匹配窗口的选择方法,与原来的矩形窗口相比较,提出的窗口选取方法能在低纹理区域得到包含有效信息量更多、更接近实际边界的窗口。此外,通过降低边界点相似度在计算过程中的权重来降低边界点影响力,使得视差不连续区域的匹配精度得到提高。算法采用了Middlebury网站上提供的四幅立体图像对Tsukuba、Venus、Teddy和Cones分别进行实验验证。实验表明文中提出的两种方法对四幅图像的匹配精度均有所提高。更多还原

【Abstract】 Two algorithms were proposed in this article to improve the matching effect of stereo image pairs.One is the improved algorithm of adaptive image window.Compared with the previous rectangle image window,the new image window obtained by the improved algorithm includes more intensity variations in low texture areas.It becomes more effective to make the new image window get close to the real edge of a certain texture area in the stereo image pairs.The other is the improved algorithm in edge pixel areas,which is used to enhance the correction rate of matching the edge pixels by decreasing the Corr(color correlation measurement) of the edge pixels.These two algorithms were investigated by matching four stereo images(Tsukuba,Venus,Teddy and Cones) with ground truth provided in Middlebury stereo database.The experimental results show that the correction rate of matching is improved obviously for all the stereo images.更多还原