Computer program reveals invisible motion in video

Login to rate this video.

You can place this video on your website by inserting the (X)HTML code below:

Embed code:
<iframe src="" width="400" height="330" frameborder="0"></iframe>

You can email this video to your friends by entering their addresses below:

Your information:

add Add another recipient

Human verification:

People who liked this video also liked

Lamb Of God - Ghost Walking (Stefanie Stuber) _ The Voice of Germany 2019 _ Blinds
Dashcam accidents compilation
The Feynman Technique
Nicholas Winton Saved 669 Children During The Holocaust...
Building the Perfect Squirrel Proof Bird Feeder
Making a GOOGOL:1 Reduction with Lego Gears


9 comments posted so far. Login to add a comment.

Expand all comments

Picture of Sustagen35 achievements

+5 1. Sustagen commented 7 years ago


Its not 'invisible'. The differences is so small that our human eyes can't detect its. But very nice technology it is. Upgrade to healthcare fields.

Picture of MrG31 achievements

0 2. MrG commented 7 years ago

Is this real? Even the glasses on the first man was "beating", the pixels on the glasses should not be changing like that? On the first baby the clothes also flashes.
And the last baby, what was the difference? On the second clip the baby's stomach is moving a lot more, that is not "invisible" moves that is suddenly revealed to the world, how can it be?
Picture of Thanny37 achievements

+17 3. Thanny commented 7 years ago

There are two completely different amplifications going on here.

For the heartbeat monitoring, they're amplifying color differences to catch the subtle color changes as fresh surges of oxygenated blood flows into the capillaries and whatnot. Things other than skin will change slightly with motion, as reflections and shadows shift.

For the breathing, they're amplifying pixel displacement to exaggerate the motion - something much more complicated, as it requires first detecting which pixels are in motion. That will work better with high contrast areas, where edge detection is much simpler.

I can see some pernicious applications here as well. Those deluded into thinking polygraphs actually detect lies might be tempted to create another fraudulent device that monitors heartbeat passively using this method, and perhaps also pick up on skin moisture changes (sample the color of incoming light and model how increased reflection of that light by sweat would affect overall color - just speculation, of unknown feasibility).
Picture of Sizzlik64 achievements

+11 4. Sizzlik (admin) commented 7 years ago

Open Source FTW :)
Picture of loadrunner54 achievements
Comment rated too low. Show this comment

-5 6. loadrunner commented 7 years ago

Would be cool in military use? then you can spot enemy snipers very easy.
Picture of 1v14k029 achievements

+3 7. 1v14k0 commented 7 years ago

MITnews article ->
I presume is the preamable research that dates back to 2005, but for the research described in this video check -> and as said the code is freely availabe, though you have to use Matlab if you want to make best use of it, and keep in mind it's under non-commercial research purposes only license.
Picture of orion27 achievements

+1 8. orion commented 7 years ago

I'm very skeptical... are they using raw video from very good cameras? Because if you take a compressed video stream or a bad camera with a lot of noise, the flicker and artifacts would overpower the microscopic sub-pixel motion.
Picture of fmohiy24 achievements

+4 9. fmohiy commented 7 years ago

ok, now lets think of something nasty we can amplify 1000 times >:)