Conventionally, an integral imaging light field display's elemental image array (EIA) is computationally generated by setting a virtual camera array. The method's high computation complexity hinders real-time rendering. Previous studies utilized high-performance parallel computing to accelerate, which, however, brought about high power consumption and cost concerning portable VR, AR, and MR devices. Some studies used a sparse camera array for fewer computations, but the generation accuracy had to be sacrificed. In this Letter, we propose a new rendering method reducing the computational complexity hundreds of times. We demonstrate that every voxel is invariably reconstructed by several homogeneous pixels on the image source. The invariable voxel-pixel mapping is pre-stored as a lookup table (LUT). Once a target 3D scene is inputted, it is resampled to a voxel array. Next, the voxel array is fast mapped to an EIA through the LUT. If a typical rectangular EIA is used, the mapping can be implemented in columns and rows, further reducing computations and the LUT size. The proposed method requires no advanced computing hardware and is independent of the model used for raytracing. Experimental verification.
翻译:常规上, 集成成成像光显示元素图像阵列( EIA) 是通过设置虚拟相机阵列而计算生成的。 方法的高度计算复杂性阻碍实时生成。 先前的研究利用高性能平行计算加速, 但是, 却带来了便携式 VR、 AR 和 MR 设备的高电耗和成本。 一些研究使用稀疏的相机阵列来进行较少的计算, 但生成精度必须牺牲。 在本信中, 我们提议一种新的配置方法来降低计算复杂性数百倍。 我们显示, 每个 voxel 总是由图像源上的若干同质像素来重建。 无法变异的 voxel 象素映射会预先存储成一个外观表( LUT ) 。 一旦目标 3D 场景被输入, 它会被重新打印到一个 voxel 阵列。 下一步, voxel 阵列会通过 LUT 快速地被映射到 EIA 。 如果使用典型的矩形 EIA,, 则可以在柱形和行中进行绘图,,, 进一步减少计算和 LUT 大小 。 。 。 拟议方法不需要用于 高级的硬件和 。