折反射全向图像柱面展开校正算法研究

Rectification of Catadioptric Omnidirectional Images Via Cylindrical Unwarping

传统图像校正方法无法直接应用于折反射全向图像。为了有效解决折反射全向图像的校正问题,提出了一种通过校正对应柱面展开图来校正折反射全向图像的方法。该文首先分析了全向图的获取与展开,而后根据柱面全景图像成像原理,讨论了其对极几何约束与求解;接着依据校正图像是否充分有效利用参考图像,提出了图像校正重采样效应的评估准则;最后分析了在这一准则下,柱面图像的基于极线采样校正方法在像素保持率和像素有效率上的优势,并重点研究了如何依据对极几何来最大限度对参考图像进行采样,以减少像素损失和图像畸变。实验结果表明,该方法可以有效提高校正图像的像素保持率和像素有效率。

A rectification approach for catadioptric omnidirectional image pair via the rectification of corresponding unwarping cylindrical image is proposed. The acquisition and unwarping process of omnidirectional image is discussed firstly. Then, according to imaging principle of cylindrical panoramic image, corresponding epipolar geometry and solving is analyzed. Meanwhile this paper develops an criteria for resampling-effect assessment depending on the degree of pixel information exploitation that rectified images based on the reference images. And it shows that, under this criterion, rectification for cylindrical image by resampling reference images along epilines has obvious advantage in terms of pixel- maintain-rate and effective-pixel-rate compared to other approaches. To reduce resolution degradation and image deformation resulted from proposed rectification algorithm, issue of how to sampling reference images maximum as much as possible according to the epipolar geometry is mainly studied and solved. Lastly results of our approach applied to both synthetic and real scenes are given at the end of this paper.