TEDGlobal 2012
Ramesh Raskar: Imaging at a trillion frames per second
Filmed:
Readability: 4
5,395,201 views
Ramesh Raskar presents femto-photography, a new type of imaging so fast it visualizes the world one trillion frames per second, so detailed it shows light itself in motion. This technology may someday be used to build cameras that can look “around” corners or see inside the body without X-rays.
Ramesh Raskar - Femto-photographer
Photography is about creating images by recording light. At the MIT media lab, professor Ramesh Raskar and his team members have invented a camera that can photograph light itself as it moves at, well, the speed of light. Full bio
Photography is about creating images by recording light. At the MIT media lab, professor Ramesh Raskar and his team members have invented a camera that can photograph light itself as it moves at, well, the speed of light. Full bio
Double-click the English transcript below to play the video.
00:16
Doc Edgerton inspired us with awe and curiosity
0
902
5978
00:22
with this photo of a bullet piercing through an apple,
1
6880
5262
00:28
and exposure just a millionth of a second.
2
12142
4878
00:32
But now, 50 years later, we can go a million times faster
3
17020
7307
00:40
and see the world not at a million,
4
24327
3596
00:43
or a billion,
5
27923
1810
00:45
but one trillion frames per second.
6
29733
3435
00:49
I present you a new type of photography,
7
33168
4385
00:53
femto-photography,
8
37553
2059
00:55
a new imaging technique so fast
9
39612
4600
01:00
that it can create slow motion videos of light in motion.
10
44212
5201
01:05
And with that, we can create cameras
11
49413
2738
01:08
that can look around corners,
12
52151
2145
01:10
beyond line of sight
13
54296
1997
01:12
or see inside our body without an X-ray,
14
56293
4393
01:16
and really challenge what we mean by a camera.
15
60686
5401
01:21
Now if I take a laser pointer and turn it on and off
16
66087
3511
01:25
in one trillionth of a second --
17
69598
2827
01:28
which is several femtoseconds --
18
72425
2812
01:31
I'll create a packet of photons
19
75237
2280
01:33
barely a millimeter wide,
20
77517
2350
01:35
and that packet of photons, that bullet,
21
79867
2926
01:38
will travel at the speed of light,
22
82793
1950
01:40
and, again, a million times faster than an ordinary bullet.
23
84743
4508
01:45
Now, if you take that bullet and take this packet of photons
24
89251
4906
01:50
and fire into this bottle,
25
94157
3173
01:53
how will those photons shatter into this bottle?
26
97330
4534
01:57
How does light look in slow motion?
27
101864
4224
02:21
Now, the whole event -- (Applause)
28
126041
3658
02:25
(Applause)
29
129699
4219
02:29
Now, remember, the whole event
30
133918
2603
02:32
is effectively taking place in less than a nanosecond
31
136521
3383
02:35
— that's how much time it takes for light to travel —
32
139904
2336
02:38
but I'm slowing down in this video by a factor of 10 billion
33
142240
4500
02:42
so you can see the light in motion.
34
146740
3673
02:46
But, Coca-Cola did not sponsor this research. (Laughter)
35
150413
4621
02:50
Now, there's a lot going on in this movie,
36
155034
2047
02:52
so let me break this down and show you what's going on.
37
157081
2362
02:55
So, the pulse enters the bottle, our bullet,
38
159443
3240
02:58
with a packet of photons that start traveling through
39
162683
2562
03:01
and that start scattering inside.
40
165245
1837
03:02
Some of the light leaks, goes on the table,
41
167082
2231
03:05
and you start seeing these ripples of waves.
42
169313
2787
03:08
Many of the photons eventually reach the cap
43
172100
2881
03:10
and then they explode in various directions.
44
174981
2914
03:13
As you can see, there's a bubble of air,
45
177895
1912
03:15
and it's bouncing around inside.
46
179807
1666
03:17
Meanwhile, the ripples are traveling on the table,
47
181473
2474
03:19
and because of the reflections at the top,
48
183947
1869
03:21
you see at the back of the bottle, after several frames,
49
185816
3634
03:25
the reflections are focused.
50
189450
2902
03:28
Now, if you take an ordinary bullet
51
192352
5894
03:34
and let it go the same distance and slow down the video
52
198246
3401
03:37
again by a factor of 10 billion, do you know
53
201647
2549
03:40
how long you'll have to sit here to watch that movie?
54
204196
5725
03:45
A day, a week? Actually, a whole year.
55
209921
4388
03:50
It'll be a very boring movie — (Laughter) —
56
214309
3914
03:54
of a slow, ordinary bullet in motion.
57
218223
4052
03:58
And what about some still-life photography?
58
222275
4603
04:08
You can watch the ripples again washing over the table,
59
232770
5352
04:14
the tomato and the wall in the back.
60
238122
2913
04:16
It's like throwing a stone in a pond of water.
61
241035
4224
04:23
I thought, this is how nature paints a photo,
62
247197
3890
04:26
one femto frame at a time,
63
251087
2587
04:29
but of course our eye sees an integral composite.
64
253674
5392
04:34
But if you look at this tomato one more time,
65
259066
3126
04:38
you will notice, as the light washes over the tomato,
66
262192
2516
04:40
it continues to glow. It doesn't become dark.
67
264708
2891
04:43
Why is that? Because the tomato is actually ripe,
68
267599
3548
04:47
and the light is bouncing around inside the tomato,
69
271147
2101
04:49
and it comes out after several trillionths of a second.
70
273248
4466
04:53
So, in the future, when this femto-camera
71
277714
2633
04:56
is in your camera phone,
72
280347
2092
04:58
you might be able to go to a supermarket
73
282439
1710
05:00
and check if the fruit is ripe without actually touching it.
74
284149
4040
05:04
So how did my team at MIT create this camera?
75
288189
5330
05:09
Now, as photographers, you know,
76
293519
1967
05:11
if you take a short exposure photo, you get very little light,
77
295486
4061
05:15
but we're going to go a billion times faster
78
299547
2206
05:17
than your shortest exposure,
79
301753
1856
05:19
so you're going to get hardly any light.
80
303609
1800
05:21
So, what we do is we send that bullet,
81
305409
1843
05:23
those packet of photons, millions of times,
82
307252
2549
05:25
and record again and again with very clever synchronization,
83
309801
3107
05:28
and from the gigabytes of data,
84
312908
2091
05:30
we computationally weave together
85
314999
2116
05:33
to create those femto-videos I showed you.
86
317115
3465
05:36
And we can take all that raw data
87
320580
2540
05:39
and treat it in very interesting ways.
88
323120
2895
05:41
So, Superman can fly.
89
326015
1841
05:43
Some other heroes can become invisible,
90
327856
2462
05:46
but what about a new power for a future superhero:
91
330318
5098
05:51
to see around corners?
92
335416
2498
05:53
The idea is that we could shine some light on the door.
93
337914
4673
05:58
It's going to bounce, go inside the room,
94
342587
2675
06:01
some of that is going to reflect back on the door,
95
345262
2430
06:03
and then back to the camera,
96
347692
1507
06:05
and we could exploit these multiple bounces of light.
97
349199
3488
06:08
And it's not science fiction. We have actually built it.
98
352687
2397
06:10
On the left, you see our femto-camera.
99
355084
2384
06:13
There's a mannequin hidden behind a wall,
100
357468
2379
06:15
and we're going to bounce light off the door.
101
359847
2982
06:18
So after our paper was published
102
362829
1948
06:20
in Nature Communications,
103
364777
1934
06:22
it was highlighted by Nature.com,
104
366711
1915
06:24
and they created this animation.
105
368626
2563
06:27
(Music)
106
371189
6402
06:33
We're going to fire those bullets of light,
107
377591
3461
06:36
and they're going to hit this wall,
108
381052
3263
06:40
and because the packet of the photons,
109
384315
2655
06:42
they will scatter in all the directions,
110
386970
2297
06:45
and some of them will reach our hidden mannequin,
111
389267
2248
06:47
which in turn will again scatter that light,
112
391515
2879
06:50
and again in turn the door will reflect
113
394394
3686
06:53
some of that scattered light,
114
398080
2072
06:56
and a tiny fraction of the photons will actually
115
400152
2744
06:58
come back to the camera, but most interestingly,
116
402896
2284
07:01
they will all arrive at a slightly different time slot.
117
405180
3746
07:04
(Music)
118
408926
4577
07:09
And because we have a camera that can run so fast,
119
413503
2817
07:12
our femto-camera, it has some unique abilities.
120
416320
3106
07:15
It has very good time resolution,
121
419426
2906
07:18
and it can look at the world at the speed of light.
122
422332
3518
07:21
And this way, we know the distances, of course to the door,
123
425850
3545
07:25
but also to the hidden objects,
124
429395
1894
07:27
but we don't know which point corresponds
125
431289
1595
07:28
to which distance.
126
432884
2322
07:31
(Music)
127
435206
3240
07:34
By shining one laser, we can record one raw photo, which,
128
438446
3944
07:38
you look on the screen, doesn't really make any sense,
129
442390
2450
07:40
but then we will take a lot of such pictures,
130
444840
1880
07:42
dozens of such pictures, put them together,
131
446720
2419
07:45
and try to analyze the multiple bounces of light,
132
449139
2678
07:47
and from that, can we see the hidden object?
133
451817
3416
07:51
Can we see it in full 3D?
134
455233
2919
07:54
So this is our reconstruction. (Music)
135
458152
2636
07:56
(Music)
136
460788
3458
08:00
(Music) (Applause)
137
464246
8308
08:08
Now we have some ways to go before we take this
138
472554
2611
08:11
outside the lab on the road, but in the future,
139
475165
3197
08:14
we could create cars that avoid collisions
140
478362
2834
08:17
with what's around the bend,
141
481196
2216
08:19
or we can look for survivors in hazardous conditions
142
483412
3914
08:23
by looking at light reflected through open windows,
143
487326
4278
08:27
or we can build endoscopes that can see
144
491604
2625
08:30
deep inside the body around occluders,
145
494229
3270
08:33
and also for cardioscopes.
146
497499
1876
08:35
But of course, because of tissue and blood,
147
499375
2501
08:37
this is quite challenging, so this is really a call
148
501876
2187
08:39
for scientists to start thinking about femto-photography
149
504063
2853
08:42
as really a new imaging modality to solve
150
506916
2589
08:45
the next generation of health imaging problems.
151
509505
3577
08:48
Now, like Doc Edgerton, a scientist himself,
152
513082
3881
08:52
science became art, an art of ultra-fast photography,
153
516963
5176
08:58
and I realized that all the gigabytes of data
154
522139
3464
09:01
that we're collecting every time
155
525603
2004
09:03
is not just for scientific imaging, but we can also do
156
527607
3443
09:06
a new form of computational photography
157
531050
3780
09:10
with time-lapse and color-coding,
158
534830
4139
09:14
and we look at those ripples. Remember,
159
538969
2770
09:17
the time between each of those ripples is only
160
541739
2675
09:20
a few trillionths of a second.
161
544414
4343
09:24
But there's also something funny going on here.
162
548757
1956
09:26
When you look at the ripples under the cap,
163
550713
2354
09:28
the ripples are moving away from us.
164
553067
3620
09:32
The ripples should be moving towards us.
165
556687
2149
09:34
What's going on here?
166
558836
1767
09:36
It turns out, because we're recording
167
560603
1958
09:38
nearly at the speed of light,
168
562561
4547
09:43
we have strange effects,
169
567108
2070
09:45
and Einstein would have loved to see this picture.
170
569178
4031
09:49
The order at which events take place in the world
171
573209
3269
09:52
appear in the camera with sometimes reversed order,
172
576478
4568
09:56
so by applying the corresponding space and time warp,
173
581046
3359
10:00
we can correct for this distortion.
174
584405
4045
10:04
So whether it's for photography around corners,
175
588450
4241
10:08
or creating the next generation of health imaging,
176
592691
4308
10:12
or creating new visualizations,
177
596999
2680
10:15
since our invention, we have open-sourced
178
599679
3559
10:19
all the data and details on our website, and our hope
179
603238
3699
10:22
is that the DIY, the creative and the research community
180
606937
6636
10:29
will show us that we should stop obsessing about
181
613573
3803
10:33
the megapixels in cameras — (Laughter) —
182
617376
3240
10:36
and start focusing on the next dimension in imaging.
183
620616
5030
10:41
It's about time. Thank you. (Applause)
184
625646
4534
10:46
(Applause)
185
630180
10205
ABOUT THE SPEAKER
Ramesh Raskar - Femto-photographerPhotography is about creating images by recording light. At the MIT media lab, professor Ramesh Raskar and his team members have invented a camera that can photograph light itself as it moves at, well, the speed of light.
Why you should listen
In 1964 MIT professor Harold Edgerton, pioneer of stop-action photography, famously took a photo of a bullet piercing an apple using exposures as short as a few nanoseconds. Inspired by his work, Ramesh Raskar and his team set out to create a camera that could capture not just a bullet (traveling at 850 meters per second) but light itself (nearly 300 million meters per second).
Stop a moment to take that in: photographing light as it moves. For that, they built a camera and software that can visualize pictures as if they are recorded at 1 trillion frames per second. The same photon-imaging technology can also be used to create a camera that can peer "around" corners , by exploiting specific properties of the photons when they bounce off surfaces and objects.
Among the other projects that Raskar is leading, with the MIT Media Lab's Camera Culture research group, are low-cost eye care devices, a next generation CAT-Scan machine and human-computer interaction systems.
Papers:
Andreas Velten, Thomas Willwacher, Otkrist Gupta, Ashok Veeraraghavan, Moungi G. Bawendi and Ramesh Raskar, “Recovering ThreeDimensional Shape around a Corner using Ultra-Fast Time-of-Flight Imaging.” Nature Communications, March 2012
Andreas Velten, Adrian Jarabo, Belen Masia, Di Wu, Christopher Barsi, Everett Lawson, Chinmaya Joshi, Diego Gutierrez, Moungi G. Bawendi and Ramesh Raskar, "Ultra-fast Imaging for Light in Motion" (in progress). http://femtocamera.info
Ramesh Raskar | Speaker | TED.com