摘要
由于融合函数中的Census变换部分的代价计算取决于中心像素点的灰度值、易受光照和噪声影响,绝对误差和SAD代价函数不能利用图像色彩信息导致精度低等缺点,提出一种使用图像色彩信息来校正像素点代价值的Census变换和有权重的SAD代价函数相结合的局部立体匹配算法。在代价计算部分,使用改进SAD和Census变换的代价值结合,形成该像素点的融合代价值;在代价聚合部分,使用十字交叉域聚合的方法,得出结果后再进行迭代,可以得到精度更高的视差图。该算法通过引入邻域像素点的色彩信息,计算出权重以及色彩校正参数,然后对中心像素点的融合代价值进行优化,从而降低SAD-Census算法的误匹配率。实验结果表明,使用了色彩校正参数的改良SAD-Census算法,有效降低了图像的误匹配率,适用于色彩信息重复度不高的图像。
Due to the high dependence of traditional Census algorithm on the center pixel,being susceptible to light and noise,being unable to use the information of image color,and low accuracy of the sum of absolute differences(SAD)cost function,this paper proposes a local stereo matching algorithm which uses the information of color to correct the cost of pixels.In the cost calculation part,the improved SAD and Census algorithm are combined to form a data item of cost.In the cost aggregation part,the cross based aggregation is used to get the new data item of cost.Then using iteration,the parallax map with higher precision can be obtained.By introducing the neighboring pixels with the information of color,the weight and the color correction parameters are calculated.Next,the algorithm optimizes the fusion cost value of the central pixel,and reduce the mismatching rate of SAD-Census algorithm.The experimental results show that the improved SAD-Census algorithm with color correction parameters can effectively reduce the mismatching rate of the images,and is suitable for the image with low color information repetition.
作者
王森
申鹏宇
詹小秦
伍周元
WANG Sen;SHEN Pengyu;ZHAN Xiaoqin;WU Zhouyuan(School of Science,East China Jiaotong University,Nanchang 330013,China)
出处
《机械与电子》
2024年第5期3-8,13,共7页
Machinery & Electronics
基金
江西省自然科学基金项目(20224BAB211005)。
