一种投影仪相机系统几何配准的鲁棒算法

A robust registration algorithm of image geometry for projector-camera system

目前投影屏的反射及几何特性决定了投影仪投影图像的显示效果。为解除投影仪对标准屏幕的依赖性,借助数码相机可构成反馈系统对显示表面的缺陷进行辐射度补偿和图像校正,而其前提即是建立投影图像面与相机图像面像素之间的配准关系。为实现这一目的提出了一种能实现投影仪相机系统几何配准方法的鲁棒算法。通过设计黑白二色编码图像、显示采集编码图像的方式并对捕获图像进行二值化以及形态学处理以定位和识别采样色块,建立几何配准函数关系。该算法在平面、曲面以及无图案与有图案等各种显示表面的投影场景中得以验证,其平均配准误差稳定在0.2～1.0个像素单位,且不依赖于色块的数量而变化,并能在10s内完成整个配准关系的建立。实验表明该方法能在适用范围以及精度和效率等方面满足相机反馈式投影仪像面配准的应用需求。

The existing projector relies on a high quality screen for exhibiting a remarkable image, which would be affected by the reflecting and geometric characteristics of the screen. Aiming to relax this confinement, the projector-camera system is then developed to compensate and correct, in radiometry, the shortcomings of the display quality on the screen, in which the geometric registration between the pixels of projection and camera images becomes a foundational and significant issue. Based on the methods of lump sampling, morphologic processing, identifying and locating the sampling patches, as well as binary polynomial fitting method, a registration algorithm was achieved. The established algorithm was tested under various scenes of different display surfaces, including both flat and curved screens, together with clean and colour-patterned screens. The experimental results indicated that the mean deviation of image registration was between 0.2 to 1.0 pixel units, regardless of the change in sampling quantity. Meanwhile, the executive time of the registration was less than 10 seconds. This sound efficiency and high precision implied this algorithm is applicable to practical demands.