With the struggle to display 2-D subtitles, it may be difficult to explain the importance of looking ahead to 3-D subtitles. Avatar is probably one of the best examples of why subtitling in 3-D space is so important. If you recall, Avatar presented English subtitles when Na’vi was spoken. It was probably no big deal to you, because it was easy to read. But the reason it was easy to read it that subtitles in Avatar had dimension to them. The depth of the text matched the depth of the main object of the scene, where your eye was most likely focused. If the subtitles in Avatar were in 2-D, it’s guaranteed that you’d remember them, as they would have been difficult to read due to your need to refocus your eyes. To insure that the subtitles in Avatar were easy to read, at least one subtitle actually moves in depth with the main character of the scene while present on the screen.
Avatar’s diligence in rendering 3-D subtitles was no accident. Industry tests of subtitles with 3-D movies demonstrated that it was necessary to render subtitles in 3-D, and that there would be times where it would be necessary to follow the depth of the main object of the scene. This spawned several concerns. First was that the current SMPTE DCP has no way of packaging the information needed to render text or graphic subtitles in 3-D. Second was the very real concern that projector manufacturers did not have the additional processing power available in their projectors to properly render 3-D subtitles, particularly 3-D subtitles that move in depth.
An important demonstration was made at NAB Digital Cinema this month by Dolby in which 3-D subtitles were rendered in comparative ways that would determine the degree of horsepower needed. Dolby was an important company to perform the tests, as Dolby’s server has little or no horsepower available for rendering subtitles, giving its engineers a very sharp focus on providing value with minimum impact to their product. The demonstration was repeated in its Burbank in-house theatre, where it was possible to observe it in a more intimate environment.
Two horsepower intensive operations were presented. Captions were presented in 3-D with aliasing, and without. Text presented on a pixilated display is subject to the blocky, jagged edges that one might see with text on their computer monitor. To eliminate this anomaly, a process called “anti-aliasing,” or smoothing is enabled, which introduces a “blur” to the edges and smoothes the appearance of the text. Dolby presented subtitles in 3-D with and without anti-aliasing. In all cases, the text with smoothed edges provided a noticeable improvement in appearance. This requires more computational power, but not as much as would be required of the second test.
The second test was of z-axis motion in 3-D space. The perception of depth is invoked by separating the left view of an object horizontally from the right view. If moving the depth of the object, the left view must move horizontally in relation to the right view of the object. A similar anti-aliasing process can be invoked that smoothes the movement of the text as the left view moves horizontally in relation to the right view. Essentially, a motion blur is introduced. Without the anti-aliasing process, the left/right view of the text is simply stepped horizontally in a pixel-by-pixel manner, requiring the smaller amount of computational horsepower.
The z-axis test was enlightening. With glasses off, one could easily see the left/right views jump in a pixel-by-pixel manner. The appearance was rough. With 3-D glasses on, over 90% of the roughness disappeared, producing an acceptably smooth movement in depth. This tells a stunningly unexpected story about the human visual system. The z-axis movement without anti-aliasing is probably not advised where the audience is exposed to a large number of subtitles requiring it. But subtitle experts do not expect that z-axis movement is needed on a regular basis. Therefore, Dolby’s discovery will prove useful to those who must minimize the amount of computational horsepower needed to render the 3-D subtitle.
From a standards perspective, this removes the concern that allowing for such z-axis movement in the standard will over-complicate 3-D subtitle rendering engines. Dolby’s discovery eliminates what is perhaps the last block in the evaluation process for standardizing the distribution format for 3-D subtitles.