通用加速计算在环幕失真校正中的应用研究

Research on General Acceleration Computation Applied in Distortion Correction

摘要:

大规模场景的沉浸式投影系统大都以环幕作为显示设备,而由环幕造成的几何失真使系统真实感大大降低。从环幕产生失真的几何原理出发,通过建立投影平面和成像面之间的坐标映射关系,生成了一种几何失真校正的理论模型。同时,研究了一种适于通用加速计算的图形硬件架构CUDA,基于其多线程的硬件执行机制,通过将几何失真的校正矩阵组织成纹理进行多重纹理贴图映射,设计了预失真校正的硬件加速算法,该算法还利用基于权值的超采样技术对加速中的走样进行了设计。最后,对该方法的校正效率和效果进行了实验验证,证明该方法具有更高的场景交互速度和很好的校正效果。

Abstract:

In large multi-projector display system, because of the nonlinear geometrical distortion, it will lose the quality of immersion. According to the geometric principle of distortion, the coordinate-mapping relation between the projecting surface and imaging surface was deduced, and the theoretical model based on geometric distortion correction was generated. Moreover, new hardware-implementation architecture was well researched, which was based on multi-thread mechanism. An algorithm used in hardware acceleration was presented by means of multitexturing, which organizing the correction matrix to a 2-dimension texture. And it also included an antialiasing design using super sampling based on weight. Finally, the experiment shows that the algorithm can enhance the interactive rendering velocity and the simulation reality.