Mimicking persistence of vision for fast-moving objects

Question

A common issue I see with videos of poi spinning is that as the poi spin faster, the video ends up with disjointed arcs of light, rather than a smooth motion as one would see in real life. Here is a frame from a video demonstrating the issue:

Notice the hard edges showing the light collected in a single frame, which to me doesn't look good. In video, it produces this jarring effect:

Example using LED poi heads

Example using fire poi heads

I want the video instead to mimic human persistence of vision, wherein each frame is blurred into the next in a way that the spinning poi will draw a smooth path.

I do know that long exposure can be used in photography to mimic the effect, but I want the effect in video. My searching online for a solution hasn't been fruitful. How can I achieve this, either using camera settings or in post processing?

Answer 1

There are two ways to fundamentally solve this, neither necessarily feasible: 1) Acquiring and playing video at high frame rates. 2) acquiring at high frame rate but using eye-tracking dependent rendering to render appropriately. The reason why is that the eye is not stationary on the scene (well, if it is, the artifacts disappear and the motion looks smoothly blurred) but rather tracks the moving lights. In this case, the lights have low speed on the retina but had high speed on the image sensor. So the image sensor had a lot of blur during the exposure although the exposure suddenly ends so there is a sharp edge to the blurred arc. If the duty cycle of the exposures in subsequent frames is close to 100%, this may not much matter because the next frame will continue the blurred arc from the edge. However, if you are tracking the movement with your eyes, you will see these sharp edges. Playing high frame rate footage at the same at a very fast frame rate (which will require a high refresh rate display) will mimic the original situation in which blurring is created due to the slow visual response in the retina being distributed over space by movement and eye tracking will allow the tracked feature to remain sharp whereas non-tracked features will be blurred.

You might be able to "cheat" by averaging frames together (e.g. a 3 frame rolling average), potentially with a non-uniform weighting. The idea would be to blur the sharp edges. If the source material is high frame rate, the blurred version could still maintain high temporal detail and not appear "too slow". This may minimize the visual appearance of the artifacts by reducing the hard edges but it won't fundamentally solve the issues created by the eye moving around the scene.

This is why action sports videographers keep the camera locked on the action - to minimize motion blur at the place where you are looking anyway.