3D Lenticular Video

As the saying goes, a picture is worth a thousand words. Building on this idea, a video is worth a thousand pictures. An animated lenticular picture can display a short video. However, can a 3D lenticular video be made? The quick answer is: YES. The longer answer is yes, but...

The simple 3D video sequence

A 3D lenticular picture is made from a sequence of shots taken from a series of equidistant points on a horizontal trajectory. A simple 3D video sequence can be created by allowing the object to move during photography. Such pictures were done, but rarely displayed. And for a good reason.

There are two methods to acquire a lenticular sequence. In one method, an array of synchronized cameras is used. In this method, the photography is instantaneous, prohibiting the acquisition of a simple 3D video sequence. The other method uses a single camera moving on a trajectory. During the camera movement, the object can move too, enabling the acquisition of a simple 3D video sequence. The viewing scenario of a picture made from a simple 3D video sequence is shown schematically in the drawing below.

Figure 1: Display of a simple 3D lenticular video

The rectangles array in Fig. 1 represents the sequence, and each rectangle represents a sequence image. In this illustration, the observer's left and right eyes are exposed to the object views at times t1 and t2, respectively.

In still 3D sequences, the object is stationary, and the time is insignificant. In such a case, the images to which the observer's eyes are exposed form a proper stereo pair. This is not the case for the simple 3D video sequence: the object states at different times are different, and the pair of images to which the observer's eyes are exposed is no longer a proper stereo pair. This may be acceptable if the movement between the two viewed frames is small. But, in most cases, the 3D sensation in watching a simple 3D video picture is severely hindered and spoiled.

True 3D video sequence

For a true 3D video display, both eyes of the observer must be exposed to frames shot at the same time. This requires two sequences, one for each eye. The first frames of the two sequences are shot at time t1, the second frames at time t2, etc. A true 3D video sequence is illustrated in the drawing below, where each sequence has six frames. The lenticular sequence is composed of the two sequences joined together. In this example, the length of this sequence is 12. A true 3D video lenticular sequence can be acquired using a moving stereo camera, as opposed to a conventional camera, as in the simple 3D video case.

Figure 2: A true 3D video sequence

Fig. 3 below shows the viewing scenario of the 3D video picture. The green dots represent the observer's pupils.

Figure 3: Viewing scenario for time t1

Let us denote the interocular distance by e. For viewing the 3D video picture, the full extent of the lenticular sequence in the viewing plane must be 2e, typically 130mm. This is usually much smaller than the sequence extent in common lenticular pictures. But if the sequence extent is adjusted to 2e, the magic happens: the observer's eyes are exposed to frames taken simultaneously, causing a distortion-free 3D sensation.

Fig. 3 illustrates an exposure to instance t1. When the observer moves to the right, they can see the object in other instances. At the edge of the viewing interval, the observer will see the object at time t6 with both eyes, as illustrated in Fig. 4.

Figure 4: Viewing scenario for time t6

Discussion

A 3D lenticular video picture must be viewed from a distance at which the spatial extent of the sequence is 2e (Fig. 3). Deviation from the designed viewing distance will spoil the temporal synchronization between the observer's eyes and deteriorate the 3D quality. This restriction is the primary shortcoming of the 3D video display, limiting its range of applications.

A 3D lenticular video picture must be made with a narrow viewing angle lens of about 10⁰. Pictures with larger viewing angle lenses will have viewing distances that are too short for comfortable viewing.

The observer can resolve the sequence images only if their pupil size is smaller than the spatial extension of a sequence image in the viewing plane. This limits the length of the true 3D video lenticular sequence. Assuming e = 65mm and a pupil diameter of 5mm, the maximal length for this sequence is 26.

Example

Let us consider using a 20lpi lens with 10mm thickness. The viewing angle of such a lens is approximately 11⁰. For such a viewing angle, the required sequence spatial extension is achieved at a distance of 680mm. This is a comfortable viewing distance.