基于双向映射的非网格型全向图展开插值方法

原始的全向图存在严重的同心圆环状变形,通常需展开为更适合人眼观察的柱面全景图像。结合折反射全向成像特点,提出一种基于双向映射的全向图展开为柱面全景图像的插值方法,通过光路跟踪得到图像间的坐标映射关系,将所要生成的柱面全景图中每一点正向映射到原始的全向图中,确定插值邻域点;将这些邻域点逆向投影到柱面全景图中,针对邻域点在柱面全景图中呈非网格型分布的特点,采用非均匀二维插值算法进行图像插值运算。该方法既准确地定位插值邻域点,又避免了全向图的非线性畸变对邻域点权重计算的不利影响。实验证明该方法能有效提高插值精度和展开后图像的质量。

Panoramic Unwrapping Interpolation Method of Omnidirectional Images Based on Bidirectional Mapping

Original omnidirectional images usually need to be unwrapped to cylindrical panoramic images according to human visual habits because of concentric annular distortion. According to the characteristics of catadioptric omnidirectional imaging, an interpolation method for panoramic unwrapping based on bidirectional mapping was proposed. The coordinate mapping relationships were obtained by tracking optical path. Firstly, each pixel in the target cylindrical panoramic image was projected backward onto the original omnidirectional image, and the neighboring pixels for interpolation were located. Then, all the neighboring pixels were projected forward onto the cylindrical panoramic image, and the color value of the target pixel was interpolated by using inverse distance weighted method or 3D-plane interpolation method due to the scattered distribution of the neighboring pixels in cylindrical panoramic images. The proposed method avoids the adverse effect on computing the interpolation weight of neighboring pixels caused by the nonlinear distortion, and the neighboring pixels for interpolation are also accurately located. The experimental results indicate that the proposed method is effective for enhancing the precision of interpolation and the image quality of unwrapped panoramic images.