Objective Holographic 3D display has gained considerable attention owing to its ability to completely reproduce the wavefront information of a 3D scene. To reconstruct a real and virtual 3D scene, computer-generated hologram (CGH) is found to be more flexible than optical holography. The CGH has two main branches, namely, the dynamic holography reconstructed by liquid crystals on silicon (LCoS) and the static high-resolution holography realized by photolithography. However, due to the large pixel size and low-resolution, dynamic holographic 3D display based on LCoS can only achieve a small field of view and small display size, thus limiting its application. High-resolution CGH can be calculated and printed using photolithography and copied for mass production. However, the computational time is a big challenge. To overcome this challenge, holography is regarded as an information encoding method, which is employed to encode light-field images, thereby significantly reducing the calculation time. Moreover, high-resolution holography can only reproduce a static 3D image, indicating that the viewer can see the perspective information of the 3D image from different angles, but with a static 3D image. Thus, more vivid holographic 3D display needs to be realized. In this study, multiview dynamic holographic 3D display is proposed by coding the dynamic light-field images fused from multiple rendered light-field images via 3D animation. In this new display, the viewer can see a 3D dynamic image when the viewpoint changes.

Methods In the hologram calculation, the sequence of the light-field image of each frame of the 3D model in the 3D animation was first rendered according to the pinhole array projection model. Then, the light-field images that correspond to the view angle information were extracted from the rendered multiple sets of light-field image sequences and fused to achieve the fused dynamic light-field image sequence. In the hologram coding, the frame of images in the dynamic light-field image sequence was taken as the object light amplitude, and the phase of the diverging spherical wave coming from the pinhole was taken as the object light phase. In addition, the plane reference light was used to obtain a unit hologram. Since the calculation of the dynamic light-field image and each unit hologram are independent of each other, we used parallel acceleration in the calculation process. Light-field image fusion and holographic coding of a high-resolution hologram with a size of 32mm×32mm and a resolution of 100000pixel×100000pixel only took 27min.

Results and Discussion A 3D model with a size of 18.8mm×30mm×17mm containing 4.06×10 ⁵ object points was employed for high-resolution hologram calculation. The distance between the 3D model and the hologram plane was 17mm, and the 3D animation was rotated horizontally, including 50 frame. The 355×139×50 light-field images of the 3D animation were first calculated, and the 355×139 dynamic light-field images fused from the rendered light-field images were then utilized for hologram calculation. The high-resolution hologram with a size of 32mm×32mm and a resolution of 100000pixel×100000pixel is calculated and printed using our holographic output system. The dynamic light-field image fusion and hologram calculation only took 27min. In the hologram reconstruction, a white LED light from a mobile phone was used to illuminate the hologram at the back of the hologram at a proper distance and illumination angle. A USB camera on a one-dimensional rail was used to take pictures of the hologram. When the camera was focused on the holographic plane, the holographic plane became clear, whereas the reproduced 3D image became blurred. When the camera was focused only on the 3D-reproduced image, the reproduced image became clear, whereas the holographic plane became blurred. This phenomenon indicates that the proposed display can realize the 3D light-field reconstruction. By moving the camera, the reproduced images of the other two perspectives were taken. Based on the results, we can infer that the 3D image actions are different in different perspectives. In the actual view, human eyes can see the continuously changing 3D animation by changing the viewpoints.

Conclusion High-resolution multiview dynamic holographic 3D display is realized based on the direct coding of dynamic light-field images. To produce the dynamic light-field images from the 3D animation, the mapping relationship between the viewpoints and the time sequence of 3D animation is established. The proposed 3D display is vivid with a dynamic 3D display from different viewpoints. The calculation method is also found to be effective. Although the display is a monochromatic 3D display, the color 3D display can be achieved using the combination of the color rainbow holographic method and the proposed method here. This will be the focus of our future work. Moreover, a large hologram can be produced for a more complex 3D animation to improve the practicality of the proposed method.