基于高斯色彩模型和极线约束的Daisy匹配方法

Daisy Matching Method Based on Gaussian Color Model and Epipolar Constraint

针对水中的光线传播易受影响造成成像质量下降以及水下立体匹配不再满足空气中极线约束条件等问题,提出一种基于高斯色彩模型和极线约束的Daisy匹配算法。首先将原始图像转换为高斯色彩模型图像,增强图像信息的鲁棒性;然后利用可快速计算性高斯卷积的这个特性,daisy算法快速提取像素点的特征描述子,匹配时间大大降低;最后引入共用折射面成像的极线约束模型,提高匹配正确率。实验结果表明,本文算法在引入共用折射面成像的极线约束模型相比SIFT算法的匹配精度更高,并且可以在整幅图上进行稠密匹配,且提高了匹配速度,成功将Daisy匹配算法应用于水下环境。

Aiming at the problems that the light propagation in water is susceptible to affect the imaging quality and the underwater stereo matching no longer satisfies the epipolar constraint in the air,a Daisy matching algorithm based on Gaussian color model and epipolar constraint is proposed.Firstly,the original image is converted into Gaussian color model image to enhance the robustness of image information.Then,using the fast computability of Gaussian convolution,the Daisy algorithm extracts feature descriptors quickly and densely,and the matching time is greatly reduced.Finally,the epipolar constraint model of common refractor imaging is introduced to improve the matching accuracy.The experimental results show that the proposed algorithm has higher matching accuracy than the SIFT algorithm under the epipolar constraint model with common refraction surface imaging,and can perform dense matching on the whole image,and improve the matching speed.The Daisy matching algorithm is successfully applied to the underwater environment.