The way you described the Fourier formula was so clear and elegant. The best description I have ever heard. Something so fuzzy became so clear. Fantastic video.
@CalvinoBear
7 ай бұрын
Seconded! Amazingly great explanation.
@JeSuisNerd
24 күн бұрын
You just effortlessly explained in minutes what others struggle to do in hours, and made it feel intuitive. What a fantastic teaching resource for Fourier transformations!
@brotherdust
3 жыл бұрын
This might be the most cogent explanation of FT I’ve ever encountered. Thank you!
@gastonpossel
3 жыл бұрын
Amazing. I've been familiar to FFT's since college (I'm an acoustic engineer), and used it a lot, but never had seen such a clear explanation before. Never seen an optical FFT before either!
@BioMedUSA
3 жыл бұрын
It is as though you have condensed and presented an intriguing novel with the lucid flow of exposition, rising action, climax and denouement. What an incredible amount of effort you invested to elevate the science into beautiful art.
@SamZeloof
4 жыл бұрын
this is so good! I love seeing the setups. thanks for sharing. I remember reading about an old government project that was trying to identify missile silos from aerial images using this technique, before computers
@HuygensOptics
4 жыл бұрын
Well that certainly is exotic. I guess they never got it to work because the recognition was not specific enough and they were hitting silos with grain and cattle food instead...
@jr5234
4 жыл бұрын
I was about to leave a comment about this too. I couldn't recall if it was cold war era silos or WWII era aircraft identification but the necessity (I assume) of lasers eliminates the later.
@theunknown4834
3 жыл бұрын
@@HuygensOptics Can you use this for ift?
@ShahidPalasra
2 жыл бұрын
They used this kind of setup to reconstruct image from synthetic aperturae radar.
@Eren-he5dt
9 күн бұрын
holy shit i never knew fourier was something more than a 'mathematica trick', truly im amazed at this
@loading...3197
4 жыл бұрын
Just when I thought Saturdays couldn't get any better! Your channel is amazing : )
@PixelSchnitzel
3 жыл бұрын
Absolutely, unequivocally true!
@jimfrazer123
2 жыл бұрын
As many others have noted, this is a wonderful video clip ! Optical fourier filtering was done on seismic data a few km from you at the Shell Laboratory in Rijswijk over 50 years ago. Research geophysicists did bandpass, wavenumber and even deconvolution filtering on a set up there. I never saw it in action but saw some of the results. The technique had a low dynamic range, was very fiddly, and was quickly superseded by digital methods, though some optical filtering research continued there into the 70's. They sent optical equipment for a system to their exploration group in Melbourne. I made a brief, unsuccessful attempt to get it to work there without a manual (or experimental talent).
@sumedh-girish
6 ай бұрын
This video made things so exceptionally clear to me that by the time I was done watching the video, my ears could calculate FFT's all by themselves... All jokes aside, thanks for the amazing content.
@ke9g
2 жыл бұрын
This is how the toy laser pointers that project images work, but in reverse. The swappable 'masks' that you change on the tip of the laser pointer are fourier images on photographic film.
@spicken
3 жыл бұрын
This is a brilliantly clear explanation! My compliments. As a side note you can transform back using an other lens, which means you can select parts of the spatial frequencies to get edge enhancement, or removing unwanted raster lines as was done by Nasa on some of their satellite photos.
@garrysekelli6776
3 жыл бұрын
That's why all of NASA's fotos are fake CGI.
@supreetpa1589
3 жыл бұрын
Sir you have accomplished what today's education and engineering schools have failed to do . Thank you and the youtube algorithm
@lbgstzockt8493
9 ай бұрын
I find the comment at 14:15 very fascinating, but not for the obvious reason. I think it is amazing that a medium speed CPU that is doing hundreds of things simultaneously is "only" a few billion times slower than a beam of light doing the same thing. It speaks to just how far modern computers and optimisations have come in the past decades, as before that time doing an equivalent calculation to a physical phenomenon would have been unthinkable.
@smirkingman
Жыл бұрын
The most limpid demonstration of FFT that I've ever seen. Amazing, thank you.
@gingermany6223
3 жыл бұрын
Really great explanation. As a photo-lithography engineer, I can confirm that your accent adds a level of authority when talking about optics to my ear! Keep up the great content.
@justin.booth.
2 жыл бұрын
I'd love to see the inverse FT of those optical patterns to see how well the match up with the original ones!
@zagaberoo
4 жыл бұрын
Absolutely incredible. I've enjoyed learning quite a bit about many different applications of Fourier transforms, but nothing like this. Thank you for sharing, and say hi to the budgies for us!
@bf0189
4 жыл бұрын
Very cool proof of concept demo at the end! Also I can't express how important Fourier transforms are and their wide ranges of uses even in places where you'd least expect it. Definitely one of the most important and useful things I've learned over the years!
@nihil_3380
2 ай бұрын
Incredible explanation of the Fourier Transform!
@rzalman96
4 жыл бұрын
Im in the middle of goodman's book, you made so much order in so many things. Amazing work, thank you.
@roderick.t
2 жыл бұрын
Since much of the energy of the Fourier transform is concentrated near the central region, one method to reduce this effect is to offset the object (letters A, B) from the optical axis of your setup. The optical Fourier transform is invariant under translation in the filter plane such that the correlation peak will also be offset. Instead of a photodiode, one can use a camera to observe the correlation spike in the image plane. If you move the object, the correlation spike will also move.
@btg837
2 жыл бұрын
I had a good chuckle when you compared the speed of the lens to that of the computer. Years ago in my college physics class, I remarked how complicated and one of our mathematical results seemed to be and how long it took to derive. My professor exclaimed "just think that Mother Nature solves this system of equations instantaneously!"
@carvoloco4229
3 жыл бұрын
Not only you look like a very nice couple (I like the way you switched voices in the end), your explanation of the Fourier Transform was so unexpectedly clear! Clear as light which, I guess, is appropriate for this channel.
@bielanski2493
3 жыл бұрын
Dankjewel, Jeroen! I now understand much better how the Arago effect plays into everyday optics and recognition. My supermarket bar code scanner is much less a magician today.
@admiraincan587
2 жыл бұрын
I can’t believe I understand now the FT space of frequencies. I just couldn’t do it before by reading papers or watching other videos. Thank you for such a wonderful explanation!!
@JustinKoenigSilica
2 жыл бұрын
.... you literally explained the fourier transform better than any of my professors ever could 10/10
@cryptoinside8814
Жыл бұрын
I aced my signal analysis class at U.C. Berkeley. I can do Fourier Transform in my sleep, but I never understood the fundamental reasonings. Your explanation is by far the best and easiest to understand. Thank YOU !!
@davidwilkie9551
3 жыл бұрын
Excellent expositions of why what everything is, is the zero-infinity integration of pulse-evolution differentiates here-now-forever, of time-timing, phase-locked sync-duration recirculation-> re-evolution. What You See Is What You Get in this format. This is real-time Quantum Operator Logic Fields Computational Information Technology. Instantaneous coherence-cohesion objectives in temporal superposition identification. A very gratifying for Physicists video who want to learn by doing experience before making up theories about what the origins of dimensionality coordination is in Actuality. Time Duration Timing Conception.., QM-TIME Universe.
@swag_designs5470
Жыл бұрын
Amazing, also never seen such an intuitive explination for the Fourier Transform
@BeatPOWERvomPowerhof
3 жыл бұрын
Best description of FT I have ever seen. Didactically unbelievable.
@YodaWhat
Жыл бұрын
Fascinating, and deeply thought-provoking!
@capnthepeafarmer
3 жыл бұрын
I went to college for mechanical engineering and we did a lot of work with signal processing and FFT for those signals, and in your short explanation was far more illuminating than my college was able to do in years of instruction.
@PCMcGee1
3 жыл бұрын
Another amazingly insightful explanation. I can't tell you what a pleasure it is to see your videos and share them.
@spamdump4459
3 жыл бұрын
Takes me back to a class on Fourier optics at university. I remember I couldn't understand what they meant by the Fourier transform in the focal plane. After all Fourier transform is just numbers representing frequencies. How can it exist in a physical plane? The moment we got in the lab and stuck some pinholes in the laser beam, I got it. Needless to say I was the most excited student in the lab that day, It proved to be one of my favorite classes of all time. Thanks for reviving that memory. b.t.w. you explained better than my professor.
@michaflak1370
3 жыл бұрын
University did not make me understand the Fourier transform, but this video did. Thank you!
@k8_lynnew-when455
6 ай бұрын
My favourite video on Fourier Optics right now ! This is going to help so much with the class presentation i have to give :)
@anthony4403
3 жыл бұрын
By far one of the best visual explanations I've seen on Fourier transforms and it's relation to optics. Thank you for posting!
@argenisg61
Жыл бұрын
I hope you got a patent and get the reward you deserve for this beautiful idea.
@michaelroyz2134
2 жыл бұрын
So far the best explanation of the Fourier transformation that I have seen. Well done!
@Maclman1
3 жыл бұрын
This is incredible. Thank you so much for posting it all!
@pdorfigliodikmer1098
2 жыл бұрын
Watched a full series of video on how scientist manage to image distant exo-planets without actually getting all the Fourier transform model gibberish. This video illuminated me
@blablabliam
2 жыл бұрын
I'm trying to learn Fourier Optics, and this was an amazing visual way to see the math. Thanks for making this!
@homo-sapiens-dubium
3 жыл бұрын
Great demonstration & explanation of the fourier transform / mechanism! Some computational perspective: the fft takes O(n^2 log n) to compute for a square n x n image. To read out the signal on a photo-sensor of the same size, it takes O(n^2) time. So the gain in time is miniscule (log n).
@drdca8263
2 жыл бұрын
Why would the photosensor have to be of a comparable size? Don’t you mostly just need to distinguish between more and less, not so much where on the sensor it hits?
@homo-sapiens-dubium
2 жыл бұрын
@@drdca8263 if youre using radio then youre right, the satelites use laser afaik, it is more efficient & possible in the vacuum of space.
@blueblimp
Жыл бұрын
The demos at 12:50 are amazing. The match is so close.
@cogoid
3 жыл бұрын
Before 1978, Fourier Optics was the standard method for reconstructing real images from the data recorded by the Synthetic Aperture Radar (SAR) systems. The received radio signals were recorded on photographic film, and then the film was developed and passed through an optical Fourier transform machine to convert the information into the image of the landscape. SAR technology is still used today in earth observation satellites to obtain high resolution images at night and through the clouds, but all the processing is now done electronically.
@andreasboe4509
Жыл бұрын
You are every bit as clever as Christiaan Huygens. Inspiring.
@desmond471
3 жыл бұрын
This is best ever video that describes Fourier so clear! Thank you so much! I have not understand Fourier for four years. You only make me understand it in five minutes! Hope you can make more videos. Thanks!
@troymeister100
Жыл бұрын
Brilliant presentation. The amount of work & preparation you put in is extraordinary, and your explanations, crisp & intuitive. Thank you.
@fmaion
2 жыл бұрын
This is one of my favorite videos in KZitem watched several times, and wish I could 👍 it every time. The idea of light processing is on my mind for a long time and after watching yours videos the ideia is trapped like loop 😀 The logic gates for light. Pattern recognition. I'm pretty sure this will be the future of computers and AI (pre trained "filters"), or ASICs. Opto Electronics will be the next technological breakthrough!
@CherkasovN
3 жыл бұрын
Excellent explanation of the Fourier transform. Well done for not going well-trodden ways!
@MissNorington
Жыл бұрын
I am excited that your information might be the very one to help me program a "realistic" star filter. I didn't expect your videos to be so easy to follow, even the most difficult topics.
@daynosdr
5 ай бұрын
thank you for all you do, this is fascinating. your videos have inspired me to build the autocollimator ive always wanted.
@daynosdr
5 ай бұрын
@huygens optics
@vladimirsch.3015
2 жыл бұрын
I understand the fourier transformation and did calculate it manually a few times, but I just couldn't fully grasp the Fourier optics. Several PhDs tried to describe it to me with no success. Books didnt help ither. Yout did a great Job in a few minutes, you gave me the right understanding for it. Thank you so much.
@shahademad9410
Жыл бұрын
I would like to thank you for the clear explanation
@steffanjansenvanvuuren3257
Жыл бұрын
After watching this video, it is clear that such advanced audio processing could not have come into existence by random chance as the theory of evolution states. It is clearly advanced bio-mechanical-electrical engineering. God's bio-technology is so amazing...
@MadScientist267
3 жыл бұрын
Good stuff man. I like the clean informative approach too. Increasingly rare these days.
@martin09091989
2 жыл бұрын
Absolute fascinating topic! People like you should teach such stuff for big money! The knowledge you gave my in under 20min would take an average school weeks! Thank you for sharing this with others!
@geekswithfeet9137
2 жыл бұрын
I’m sending this video to friends just to explain Fourier transforms, that cut right through the usual bs everyone else goes through to make themselves feel smart (which usually alienates the audience)..... this just felt like it was accessible to a 10 yr old, and still have a decent understanding
@tiggerbiggo
3 жыл бұрын
I'm actually speechless... I've been trying to understand how to implement FFT in a computer for so long, every explanation i've found has eluded my comprehension. And you just made me get exactly how to program it with some green and red shading on a graph. I love this channel, thanks so much for this video!
@inv41id
3 жыл бұрын
It is indeed a really easy to understand explanation for how the Fourier transform works, but the *fast* Fourier transform algorithm is significantly harder to do. So sadly you didn't learn how to implement FFT, you only learned how to implement a naive slow FT.
@tiggerbiggo
3 жыл бұрын
@@inv41id ah yeah i understand, it just so happens that the naive method is precisely what I needed because I don't care about phase for my applications.
@inv41id
3 жыл бұрын
@@tiggerbiggo Oh I wasn't really talking about phase, I was more so saying that the naive method is much slower, which is pretty significant considering FFT has the word "fast" in the name
@tiggerbiggo
3 жыл бұрын
@@inv41id yeah that too, at least my understanding is a bit better than it was before, and this naive method seems like it would still be useful in smaller applications where speed is not a big factor. Definitely not suitable for use in audio effects because of the speed and lack of phase information which would be critical for doing the inverse calculation accurately
@daverei1211
3 жыл бұрын
Great video thank you. I remember hearing about an optical lens to detect planes. Now I understand how that was done, thank you.
@congchuatocmay4837
Жыл бұрын
The fast Walsh Hadamard transform is a lot simpler to compute than the FFT especially using integers. With a specialised IC hardware design even extremely fast optics would have a difficult time competing. And these fast transforms have a matrix equivalent. The FFT can be viewed as having a 2 matrix equivalent, one of sign waves, one of cosine waves. Neural networks have weight matrices. With a little bit of technical creativity you can use a fast transform as the weight matrix for a neural network with a great deal of time saving over computing the large weight matrix multiplies in conventional artificial neural networks. Eg. SwitchNet, SwitchNet4.
@meh5647
7 ай бұрын
This is the kind of stuff that makes me wish I did optical engineering instead of comp sci.
@YSoreil
4 жыл бұрын
Reminds me of the DSP classes I took in university. A real headscratcher at first. I don't think we ever applied it to image recognition but we did go in to image filters like edge detection. Goed bezig ouwe ;)
@WildEngineering
3 жыл бұрын
ive been addicted to your videos lately
@KaliFissure
2 жыл бұрын
Amazing how we have a physical high resolution comb filter in our ear. Great series of experiments into optical processing. A complexly etched surface seems to me could calculate a range of complex functions depending on where input is fed and where output is measured.
@PavlosPapageorgiou
3 жыл бұрын
I did some undergraduate physics and loved this experiment. Still amazed it's possible, compared to a digital computation.
@calvinkielas-jensen6665
3 жыл бұрын
Randomly came across one of your videos yesterday and have been watching a few as a result. You have fantastic information and you're able to explain it in a very understandable way. Thank you very much for sharing! I look forward to watching more!
@joshuazhang4910
3 жыл бұрын
Whenever I opened a Huygens Optics video I can learn something from it.
@nitinmalapally
3 жыл бұрын
Such an intuitive and simple explanation, excellent content! You could even have combined this topic slightly with convolution/cross-correlation without causing too much divergence
@cloudgalaxy9231
2 жыл бұрын
You explain everything so clearly. Every video you make is a gift to the world.
@maciekwar
2 жыл бұрын
what a cool way of showing how video compression works :P
@Digalog
3 жыл бұрын
Also explains a bit how Fourier transform is used for hashing in cryptography :)
@philorkill
2 жыл бұрын
I am amazed by your deep knowledge of the subject. I am humbled and thankful for your contribution. Thank you for sharing!
@smallcursed2328
3 жыл бұрын
This is the coolest thing I have ever seen in my life
@pentachronic
3 жыл бұрын
One would assume the cochlea works in a similar way to a mass spectrometer except it,s sound waves hitting the surface vs atoms.
@xbronn
3 жыл бұрын
boy did i enjoy this
@physicsforsome1290
Жыл бұрын
Nice video. For anyone want to do this on a low budget then read on... You need a laser pointer pen, a convergent lens, a caliber and a microscope crossed grid slide available from Amazon. Due to the poor beam quality of laser pen it serves as a good parallel beam for illuminating the crossed grid target. Place the crossed grid at the focus of the lens to observe the diffraction pattern formed with the laser. At a viewing screen some distance away from the lens observe the image of the crossed grid. Place the jaw of the caliber at the location of the diffraction pattern to observe the effect of spatial filtering. Rotate the caliber and you would be able to observe the vertical and horizontal grids appear and disappear due to spatial filtering of the high frequency components.
@occamsrazor1285
3 жыл бұрын
Oh wow....Fourier transforms....the same kind of concept (multiplying by the frequency your're looking for) is the same sort of principle behind netmasks....cool
@anatolesokol
Жыл бұрын
I would like to see an experiment where you convert the FT plane back to original image, after manipulating it a bit, like high pass filter, that should be very interesting having an optical device with no electronics that could outline objects counters...
@Muonium1
3 жыл бұрын
Exceptional channel, immediate subscribe. I would be willing to bet literally any quantity of money that this is the exact technique the National Reconnaissance Office and or CIA was using to analyze the reams of Corona spysat film in the 60s and 70s (for finding and roughly "counting" eg. jet aircraft in a particular set of film images) long before digital computational power was anywhere near sufficient for such tasks. It is effectively an analogue computer for image analysis that operates at, as you note, fantastic speed, and with extremely low energy requirements.
@ivankudinov4153
8 ай бұрын
This is an outstanding video, and a joy to listen to. Thank you very much
@brodysnook1231
4 жыл бұрын
Really enjoyed this, thanks for taking the time to make it!
@glentyan2505
3 жыл бұрын
First class explanation of a very difficult to understand subject, well done and thank you.
@yeong126
2 жыл бұрын
Nice video! I like how you used a DLP projector to simulate filters. To distinguish patterns like A from AB, one can set each filter's threshold value instead of setting one threshold value.
@n1352-m1i
4 жыл бұрын
thank you for this very clear and didactic introduction to Fourier Optics convolution. For the sake of completeness though, a hint of the difficulties to make a useful application of that lens property could have been mentioned, as if for decades the graal of the optical computer has been promised by researchers around the world, it is still very difficult to get anything more than a toy proof of concept out of an optical bench...
@HuygensOptics
4 жыл бұрын
You are right that it's not easy to make OFT work and in such a way that it has advantages over other technology. However, I think that these long standing promises might become reality soon, as companies have developed serious products lately: www.globenewswire.com/news-release/2019/03/07/1749510/0/en/Optalysys-launches-world-s-first-commercial-optical-processing-system-the-FT-X-2000.html
@n1352-m1i
4 жыл бұрын
@@HuygensOptics I know about this accelerator, I'm still waiting for an independent benchmark to figure out exactly the level of performance it actually offers...
@toriknorth3324
3 жыл бұрын
In addition to the positive filter you could have a negative filter. While the positive filter does find the frequencies associated with the letter A (for example), it doesn't check for frequencies that A doesn't have. In other words, the positive filter looks at the pattern to see if there are any A's and the negative filter looks to see if there is anything other than A. My reason for bringing this up is that I was wondering how to discriminate between an I and some other pattern that had an I embedded in it (which most letters do).
@pissfilth
Жыл бұрын
That Fourier Plane makes sense after seeing this video. Quite interesting.. Fourier keeps surprising me with the great number of applications of his work. Can you imagine that he had this all in his mind, in an era without electronics.. Leibniz, Euler, Fourier, Gauss (and the rest!) too. What would they have come up with when they had access to computers?
@basilal-jaml1981
5 ай бұрын
having the same spectrum while changing letter position is because Fourier transform is locality in-sensitive
@abcrtzyn
3 жыл бұрын
Very cool stuff. I would love to see something like this used in practice or commercially
@0MoTheG
3 жыл бұрын
Back in the day these setups were used to develop SAR images.
@Pillowcase
3 жыл бұрын
This is kind of mind blowing
@Scrogan
4 жыл бұрын
As a physics student who has studied Fourier transforms in at least two papers, your simple explanation really takes the cake! I’ve wondered if it’s feasible to make an analogue oscilloscope that modulates brightness instead of the y deflection plate, feeds that brightness modulation into a Fourier optical system, and displays that on a conveniently visible screen. Latency-free Fourier oscilloscope! Might want to figure out how to turn the resulting brightness modulation back into a wave in the y-axis however, possibly requiring another image-sensor and a screen, which could certainly get clunky.
@HuygensOptics
4 жыл бұрын
Many analogue scopes have a Z-input for controlling the beam intensity. I'm not sure your idea is feasible though, because the Fourier math is based on diffraction and that would be hard to realize on a phosphorescent screen.
@corsonforcas
8 ай бұрын
Fourier doesn't feel that scary now, thank you in kind
@Nicho2020
6 ай бұрын
An excellent video, thank you! I would only ask about practical times of performing the electro-optical matching process, as it must be limited by sensor self noise. I mean, there must be effectively an exposure time that is likely to be much longer than the propagation time.
@joshhennen
3 жыл бұрын
the end was great! I almost pissed myself I laughed so hard!
@miklov
3 жыл бұрын
Thank you for sharing! This was quite inspiring!
@anoimo9013
3 жыл бұрын
very good explanation of a complex but ubiquitous and practical subject. Couldnt follow the explanation of a cammera focused to infinity though
@nicktohzyu
3 жыл бұрын
if shor's algorithm uses fourier transform, could we do prime factorization with optics?
@Rollinsonn
3 жыл бұрын
Bump
@DanaTheLateBloomingFruitLoop
3 жыл бұрын
We need answers.
@drdca8263
2 жыл бұрын
I don’t think so, as, the quantum Fourier transform applies to like, the state of a collection of qubits, where the amplitudes of different possible states are playing the role of the different numbers in a usual Fourier transform, And like, there are 2^n of these numbers. I don’t yet understand well enough to say “definitely no that wouldn’t work”, but I do think I understand well enough to say that I really don’t think it would? Though.. [what follows is a confused attempt to identify more precisely what wouldn’t work by trying to think of how it would work if it did work] Ok, suppose that we split (in how we consider it, not physically split) some interval of space for the beam into 2^n equal parts, where the intensity through each part is the number. Uh, the discrete Fourier transform then, would, uhh, how does that compare to the usual one which this computes? Ok uh, so I think you would need to, change the phase of the light in a way corresponding to, exponents of a given integer modulo the number you want to factor, where the power the number is raised to corresponds to the position in the beam??? How could that be done? If the exponentiation was done by repeated squaring, Hm. Ok, I don’t know enough about optics to really say, but that doesn’t seem feasible to me? Idk
@nicktohzyu
2 жыл бұрын
@@drdca8263 I myself am not an expert, but I don't think your reasoning is correct. Firstly, your argument should also apply to electronic quantum computers, which we know do likely work. Fundamentally though, there is no "2^n distinct states" nor a need (or possibility) to split them in space. Rather, the wave of a single photon has a probability density function across these states (and phase is important too). I don't see any reason that optics cannot propagate these wavefunctions just as electronic quantum gates do. Please do correct me if my reasoning is wrong, I too want to learn from this!
@drdca8263
2 жыл бұрын
@@nicktohzyu oh, sorry, yeah if you are using n entangled photons, or whatever, uh, well, Shor’s algorithm can be done with the qubits implemented as photons. What I was talking about was, trying to implement it without using any quantum mechanical properties of a beam of light, and just using wave properties.
@turun_ambartanen
3 жыл бұрын
This video is so information dense, I love it
@Bigman74066
2 жыл бұрын
Excellent video and a very interesting subject that I had never heard of before! I hope you will be able to make many more videos like this one in the future. Respect!
Пікірлер: 275