Martial Arts Footage

In July 2011, Tessive was invited to participate in a series of motion studies at Shepperton Studios, UK.  The Director, Keanu Reeves, and the Director of Photography, Robert Shacklady, wanted to find out how the Tessive system would render a martial arts sequence compared with a standard setup. 

The interest in this test is that it examines how human motion is rendered by motion picture cameras, and how reducing temporal aliasing affects the perception of that motion.  Previous tests concentrated on mechanical systems or repetitive motions where the effects of temporal aliasing are obvious.  The look and feel of human motion when aliasing is reduced hadn't been studied previously. 

The camera used for the test was an ARRI ALEXA Plus.  In the first set of 24 fps tests, the fight sequence was repeatedly recorded by the ALEXA without the Time Filter and at different shutter angles (172.8, 90, 45, 22, 11), and then with the Time Filter fitted to the front of the lens.  For comparison here, we'll concentrate on the naked ALEXA with a 172.8 degree shutter compared with the ALEXA with the Time Filter.  

For a better view, please request a free demo footage Blu-Ray disc.  Or you can download the footage from here.

The difference in motion rendering is most apparent in the middle of the fight, when the artist's arms are nearly filling the field of view and moving rapidly. These stills show the difference in the instantaneous rendering of the motion. The conventional shutter shows the typical abrupt start and stop to the motion blur, while the Tessive filtered version shows a smoother transition into the blur.  

The second image features motion with a staff. The areas of most interest are the brass fittings on the staff, which show the abrupt transitions in the conventional shutter. These stills also highlight an interesting difference between the camera with and without the Time Filter. The Time Filter overall has a slightly blue cast due to the polarizers employed in its operation, but this was easily balanced in post. The clothing of the artist, however, exhibits a marked color difference even after color balancing. Interestingly, the actual clothing was black and grey, and had no brown coloring. The red cast on the clothing in the conventional image is due to infrared leakage interacting with the CMOS sensor, while the Time Filter cuts well into the near infrared, resulting in a more accurate image.

 

In a final, lengthy fight sequence, the Time Filter was exclusively employed. For this sequence, the ALEXA was overcranked to 60 frames per second. The Time Filter was synchronized at 60 fps with the ALEXA through the ALEXA's RS port (the RS port also provided 24V power to the Time Filter). In our presentation of this footage, the majority of the footage is presented in realtime at 24 frames per second, which is achieved by skipping frames. This results in a somewhat jittery playback for two reasons: 

1) Each frame of the sped up footage covers a very shortened time with respect to the playback rate. The Tessive exposure window function only covers 1/60 second, instead of the typical 1/24 second. This effectively reduces the antialiasing capability of the system, but some antialiasing is still apparent.

2) The ratio between 24 and 60 frames per second isn't an even multiple, so every other frame is slightly out of position in time.  

The advantage of this approach is evident when the action is to be slowed. During significant moments in the action, the playback is changed to present all frames, resulting in a normal overcranked look. In these situations, the Tessive filter nicely matches the framerate, resulting in exceptional smoothness and clarity of action. The following stills show some of the graceful motion blur applied to this sequence.

 

Tessive is extremely grateful to the people who produced this footage and for the opportunity to demonstrate the Time Filter to some extremely gifted people.  We are also grateful for their permission to share these tests with a broader audience of artists.