The Crazy Computations Inside Your Smartphone Cameras
ฝัง
- เผยแพร่เมื่อ 22 ธ.ค. 2024
- I recently got an iPhone 13 Pro. I love it. For me, the thing that sticks out most - literally - is the camera. Virtually all of today’s top smartphones can take images on par with anything you can get with a standalone.
How does your mobile phone camera work? And how did it get to be so good? In this video we will look at the amazing computer and semiconductor engineering that goes inside this impressive feature.
Links:
The Asianometry Newsletter: asianometry.com
Patreon: / asianometry
The Podcast: anchor.fm/asia...
Twitter: / asianometry
Like and subscribe. Yada yada. Check out other deep dives here: th-cam.com/play/PLKtxx9TnH76RiptUQ22iDGxNewdxjI6Xh.html
I believe that the blurred background of the cup is a normal effect in low light conditions for all cameras. It’s due to the depth of field becoming shorter with a lower (numerical) f-stop.
The conclusion is a masterpiece ! Thank you !!
The MEITU image at 9:24 -9:30 is funny!
My mom said the same about my voice!
Maybe it's some biological explanation here :D
Listening to voice of their kids makes mothers very calm and relaxed.
Very good "food for thought" that people rather take unreal, but nice looking photos than the real ones... this give us back the answer that people prefer fake than reality.
I’m glad you pointed out something I’ve been saying for years: you’re not seeing what the camera sees, you’re seeing what the phone generates. It’s not a bad thing, because the raw data coming from that sensor is close to unusable. Optical designers can’t overcome the laws of physics.
The video puts it so well! food for thought, man.
That's not quite true... Many phones nowadays allow the saving of for photos in raw format, which does bypass the image processing pipeline of the phone. Sometimes entirely, sometimes only mostly. Either way the image you get from that is *far* from unusable. Often you can make these raw format photos look leaps and bounds better than the standard processed ones. If you just let the phone to its thing you have 0 control over colours, white balance, denoising, lense correction, etc.
The phone just guesstimates some appropriate values and if it guessed wrong, well tough luck. Usually it guesses decently, but it can't beat full manual control.
Now obviously none of that matters to the average person just trying to take a photo for Instagram or Facebook or whatever, and most people who want nice and accurate photos will use a DSLR, but sometimes you don't have a DSLR with you but still want to take a good photo.
Long winded way of saying: no, raw camera data is very good actually and if you really value quality of the photo over ease of making it then you'll choose the raw data over the processed image
Your brain actually does the same thing. What your eyes see is not what you perceive, the eyes data also gets quite heavily processed.
@@insu_na Let me give short winded retort, no, small sensor raw data will never match larger sensors. It's the physics. Whatever they figure out for them would just be multiplied into larger sensor DR. How many photon-bumped electrons can a cell camera count versus a large sensor camera? You're right, one would always choose RAW over anything else, but the fact remains, RAW "what"?
@@MaxoticsTV Did you even read what I wrote? Did you see the part where I said that DSLRs make better images? Should you maybe repeat school again?
Well, photographers have been altering their images since the beginning. Filters (darken the sky / brighten the sky), tilt-shift large format cameras, toying with exposure and film speed, ‘pushing’ the negatives, dodge and burn, retouch, etc. Now we just get fancy machine learning pocket supercomputers to do it for us.
The "problem" is the shifting of control over to machines. If everything is done by machines, then what remains human of the image? Producing negatives is an artform and so is (arguably) using Lightroom or Photoshop for digital images. If everything is done automatically by an algorithm, nothing but the scene/subject is unique anymore. Naturally, it becomes easier to take pictures for everyone. And that's a good thing. But just like painting and other expressions of art, photography is a skill to be learned and nurtured over multiple years. Perfection through machines is not the same as perfection through practice.
@@acidtears I never said their wasn’t a problem. Just the impulse to ‘fake reality’ has a history, culturally as well as technically.
Also, the same type of arguments you’re using here (traditional photography contains the traces of the artist, AI does not), are the same class of arguments used _against_ photography in favour of the traditional “arts” (ie drawing and painting).
@@acidtears Well said!
It’s about real time with software and digital assets
It's tool, medium.
You could argue quite similar for audio over video, books over radio, storytelling over books, Cave paintins over...
It's on you, what and how you use it.
This story struck a cord with me since I have been working on this stuff for 20 years. You nailed it. Just like our eyes and brains, the electronic cameras are starting to utilize all the light that is hitting the phone at any given moment. In the past, cameras just take a momentary sample of the light. But human eyes are doing it constantly...we don't have shutters in our eyes. I love where photography has gone and is going.
Man, I've been watching your videos since I am working on EUV processes and really think each of your videos is like a PhD thesis in visual form. They are so informative in a way that reading isn't. Despite your self-deprecation, I can sense that you are a deep, intelligent individual.
Your humour appeals to me. You have a dead-pan delivery that is geek-unique.
It needs to be remembered that all the advanced computational processing that happens with smartphones always trickles down in to the bigger DSLR/mirrorless cameras, when also coupled with larger more "optically" superior lenses, full size cameras will always be better. but having said that, modern smartphones are "good enough" for most people
You have a point, but there's a hard limit to how "good" an image can be. From a physical perspective. Arguably cameras have already surpassed that point and further manipulation is simply to please the viewer. A DSLR isn't necessarily technically superior, but you may find its rendered image more pleasing.
@@Mia-ln1zs err no. The whole point of improving image quality is to improving a more pleasing render.
The physics of having a small form factor like a smart phone, will always be technically inferior to a bigger format camera.
That's the whole reason why a front camera recorded TikTok video will never look like a properly produced talking head studio recording. Even tho both is just basically the same content as the subject is basically talking to the camera
The "hard limit" is still a long ways away.
Sure there will be a point where the smartphone's output is good enough to be consumed, but that's hardly a hard limit. Because at that point of time, the bigger camera's output will be still pushing the limits on what's possible. So it's forever a never ending chase
The performance focused chipsets in modern phones will always Trump that of the latest dslrs, sure they don't have the hardware when compared to the more hand held cameras BUT that being said, smartphones are designed from the ground up to be your primary device for almost everything and in that regard the cameras are also advancing rapidly might i add!
Now id love an amazing DSLR but when do we most feel like taking pictures? When we plan it, or simply in the moment.
Then for the moment having such a powerful camera in our pockets is simply mind blowing!
I love the advancment of our little rectangles and cant wait to see where they head! Likely AR glasses...
I have yet to see a DSLR or mirrorless camera that does the image stacking smartphone do, I really wish pocket cameras would do it, because stacking alone produces incredible results. If TH-cam would allow me to post links I've got a comparison between a modified smartphone with a nearly 1 inch sensor happily keeping up with a Ricoh GR III!
@@robinrai4973 There was ones cameras with full out smartphone features like screen and all. Or close to it. What we need are phones that are a bit thicker and even maybe have a camera bump to boot. That basically is a blend of a camera and a smartphone. Like yea maybe the phone is to big to have in your normal pocket. But like it still can be totally a thing one want to carry around. Like keeping the size down from a XL phone and just add extra thickness should totally be still mobile and pocket friendly enough for most people. Handbag and jacket/hoodie pockets are a real thing.
15:54 iPhone 13 Pro has a bigger sensor and lower aperture compared to the regular 13 and most other phones. I bet that's why you can notice more shallow depth of field, not because of AI work. Correct me if I am wrong but I think algorithms do not mess with background blur in regular mode. Nice video btw, good compilation of history! It's very interesting topic that deserves more attention.
Exactly this. It's just physics doing its work here because of the much bigger sensor and also much bigger aperture - not some software
I came here to say this thank you =)
As a professional photographer (who admittedly doesn’t know much about phone cameras) I immediately knew that wasn’t totally correct when he said that. Even my iPhone 8 plus achieves that shallow DOF when taking photos of close subjects with a low f/stop.
In my humble opinion, the bokeh you were talking about with the iPhone 13 pro wasn’t a case of “Portrait mode” leaking into normal photo mode, but a result of the camera sensor size being much larger than previous sensors.
My iPhone 8 plus is very capable of “blurring backgrounds” with a close up subject, especially when the camera app is telling me it’s taking the photo at f/1.8. That’s really just optical physics at work, wide aperture + close subject = less depth of field on non-focused subjects.
When he said this I paused the video and quickly snapped a photo of my water bottle close up just to make sure.
Sensor size doesn't have much to do with bokeh effect, it's the wide aperture that creates a natural looking depth of field.
After all, a full frame DSLR doesn't create a shallower depth of field than an APSC or micro four-thirds with the same aperture lens.
@@salamander405 Exactly. The camera on the iPhone 8 is tiny (roughly 17.5 sq-mm), even compared to micro four-third (~250 sq-mm) or one inch cameras (~125 sq-mm), yet it creates natural bokeh effect in close up shots because of F/1.8 aperture.
One of the best overview explanations of this field I've seen, good stuff.
Great job! Most non-photo nuts flub much of it. Many in the comments are confused because you didn't hammer home this--cell phone cameras USE COMPUTER PROCESSING TO MIMIC WHAT YOU'D GET FROM A REAL CAMERA. Yes, I know you said a few different ways. I agree. They do an amazing job! So if you want to take a photo with your iPhone 13 that looks like something you took with a $3,000 camera then you will not be disappointed.
But that photo is a computer generated image using very noisy image data from your phone's camera. It's like having a street artist paint a realistic portrait from your polaroid (in my day)
Hard-core photographers aren't interested in whether something looks "professional" or not. They are trying to focus on a specific visual truth. You can only get that through exact focus (which is hard to do manually on a phone). Natural optical bokeh (impossible on phones). Then there's exposure. Again, too time consuming on phone. Phones will not replace large cameras because the sensors are too noisy in most light--full stop.
You should have covered curved sensor development btw. Also, Foveon technology is worth exploring. It doesn't use filters. I would have said sensors are color blind in your video, but that's quibbling. If you're reading this I LOVE your stuff, pass it on to anyone I know who has an interest.
Also, Samsung pretty much left the large camera business even though they developed cutting-edge tech and had a cult following. Definitely worth a video from you!
Thanks
The quality of pictures from modern phones is crazy given the size of the hardware, thanks for unpacking some of the features which it relies on.
Thanks!
Good vid. It seems camera sensors themselves haven't improved greatly in the last few years. Or at least not nearly at the pace they were previously. For example, a great low light camera, the Sony A7SII to A7SIII did not see a huge improvement in low light performance despite a five year gap between releases. Maybe I am expecting too high a pace of improvement for digital sensors and they won't ever be able to compete with image intensifier tubes. For dedicated cameras I understand wanting personal control rather than computer control for taking shots but I do hope the big manufactures begin taking more advantage of computational photography in areas where a photographer can't easily improve it by manual controls- low light, star tracking, focus stacking.
Most optimizations are done on speed, pixel decity and the ability to handle special scenes (like requireing HDR). The CMOS technology is near its physics limit als noise is so low and the 'bucket size' of the pixel is difficult to increase (they call this technology deep trench isolation). Anyway, hopefully SPAD technology will better the sensitivity and thus performance.
@@hobbykip thanks for the insight. I wasn’t sure of the exact tech and what to expect for the future but did notice the “stagnation” of CMOS. Will look up SPAD.
A7S is primarily video camera
@@krzysztof6123 I understand that. Everything I said still stands
I think that low-light mode is pretty much a gimmick. I say this based on photos I've seen, not on personal experience. There is an art in using a digital camera and different lenses to capture as much light as possible. However, at the widest aperture and highest iso, things should stop becoming brighter. If you want to enhance the brightness further you can always go into Photoshop or Lightroom, but the image taken by the camera remains (in my opinion) a far better representation of reality.
Low light is difficult to shoot and I do think that certain improvements should be made to the imaging processors. But as it stands, I prefer to shoot "RAW" or even JPEG on my Fuji than artificially blowing up all aspects of my image for the sake of - people wanting it? Light is the most important aspect to photography and for casual users, using a ML algorithm to blow it the fuck up might be the best way to go. For anybody else, time and patience, persistence and determination - those are the key factors to "winning" in low-light and producing great pictures.
People don’t realize how much the image quality in smartphones is LARGELY due to the high amount of pre and post processing and smart optimizations. The sensors themselves aren’t that “good” especially when compared to classic full size sensors.
Yea so? Who cares?
As long as the photo ends up good what’s your problem?
@@chengong388 i didn’t mean it in a bad way, I find it amazing that smartphone can accomplish so much with so little.
@@chengong388 they're somewhat fake photos ?!
did you see that feature in Huwawei's smartphone (I guess) when you take a photo of the moon in night the moon looks so good and so real because of AI pre processing
@@chengong388 I mean sometimes you want to capture reality rather than what a company has decided looks best. Of course you can shoot some sort of RAW in many phones now but still. Or for example sometimes the iPhones portrait mode looks good a first glance but the bokeh looks totally fake when blown up to desktop (I'm sure this will improve with time)
@@shadowshadow2724 that example is totally different from what we're talking about.
Please, please, please make an episode about the manufacturing process of all those components!
How are this Bayer color filters produced so small, and aligned perfectly? How are zillions I'd super precise micro lenses made? How does optical imagine stabilization work that quickly and precisely?
DITTO!
I am also interested in the manufacture of the lenses. This is an aspect of the technology that is often overlooked.
As an retired EE it amazes me how much functionality is being packed into such a small package. Until this video I had no idea how computational intensive digital imagining was. I assumed (know what they say) most of the processing cycles were spent on compression to save on storage space.
15:46
That's actually natural bokeh on the glass there, the same as on a DSLR or mirrorless camera. It only pops up on the glass because it's so close to the camera.
You see more and more of it nowadays because of the physically larger, and subsequently more telephoto lenses to match in modern phone cameras.
Yeah, I can get very similar bokeh with the main camera on my Galaxy S22 Ultra (though I'm not super impressed with the quality of the bokeh in the pic in the video, tbh).
What's interesting on Android these days (not sure about iOS) is that you can add blur to _any_ photo, not just photos taken in portrait mode (and for that matter, they don't even need to have been taken on the phone's camera!) They do "depth from ML" segmentation on the image, and just apply blur based on the depth at the selected "focus point", and virtual f-stop. The Light L16 could do the same thing, but in that case the depth map was from stereo, not from ML.
@@PeteC62 In my opinion, I prefer real bokeh just because I can be guaranteed of the authenticity, there can't be any "error" to it, besides an out of focus shot.
@@dominicus9891 I totally agree. The quality of artificial bokeh depends on so many factors, like the accuracy of the depth map, and the way the blur is applied (low end software will just apply a simple gaussian blur, whereas better software will try to mimic the shape of the aperture blades, at the cost of slower processing). It's fine for making the subject of a portrait stand out from the background, but you're unlikely to get great bokeh balls...
@@PeteC62 There'll be a time in the near future where I believe the depth map itself will be doing the brute of the work to mimic the effect, instead of the "edge detection" which is needed now.
Things get even crazier when you realize the processing is so fast the same tech can be applied to video (4k60fps) as well...
The trouble is that the lenses are subpar if you are comparing them to real cameras. You get used to looking at cell phone pictures and they started to look good until you go back a camera with a real lens and suddenly all those that see the photograph are stunned by the detail, the depth and the accuracy.
And even with all that, I have grown a particular distaste for the kind of processing smartphone manufacturers do these days, and the thing is, people love that. I guess I am just not like the majority of consumers who never zoom-in to a picture, or ones who complain about over sharpening and what not.
Really good ! A couple of other fun ones where computation is key:
1. simulating more zoom levels/smooth zoom from multiple fixed levels and parallaxed sensors
2. using these parallaxed sensors to complement depth sensors in deeper fields where dedicated depth sensors fail
3. enhancing low light performance of the "zoomed" sensors by using the wider ones
4. using multiple sensors to calculate "field flow" to correct shake and pan blurs
5. On that note, using multiple sensors in the video domain to simulate filmic effects like focus pull
Just so many things one you have a matrix of sensors and a crap load of CPU/GPU !
One annoying marketing thing Apple do - some "exclusive" features of new phones COULD work on earlier ones, and sometimes quietly sneak into subsequent software releases on older phones (Bokeh being a prominent example)
Not sure where you've gotten this definition but a zoom lens doesn't usually take you from a long shot to a close up. Long shots are taken with long lenses (a long focal length), and close ups can be taken with basically any focal length. Zoom means the lens covers multiple focal lengths all in one.
Another thing to note is that you can actually use a smartphone camera to get a subject in focus and blur the background - naturally that is, without using any blur/bokeh/portrait modes. However that's only possible with small subjects in close-ups.
bro you are so good at explaining yourself, you are very listenable - i usually watch the pictures too but i can understand it all without looking usually. ill check out the podcast no worries
What’s crazy is how you keep finding such interesting topics to cover dude 👍🏾
Great video about the technology of cameras and smartphone cameras in particular! We also did one just now with my hand (lol) but in the video you can see how intricately and quickly a Samsung A22 focussed and quickly changes the depth of field all automatically!
What about the LiDAR feature on the iPhone 13pro? Agreeed that the computation “improvements” are a bit flawed philosophically but LiDAR is truly new 3d data we have to work with.
Very Well done!! I only get a glimpse at the how the image processing is conceived and executed but I can concur with your video here. It’s is astounding to see how amount of thoughts and executions going behind the scene of each selfies
Photography means writing with light, which seems intuitive when light is projected onto film. Projecting it onto a CMOS sensor through a matrix of color filters adds complexity most people would rather avoid, so the word remains the same. Thanks for the explanation.
When you point the phone at the relatively close object like in the minute 15:48, it's not leaking portrait mode, it's natural bokeh of the camera because iPhone 13 Pro has a bigger sensor and f stop than the previous iPhones
One other way zoom can work in addition to those you covered is with help from computational photography. Because your hand moves a little no matter what, by taking multiple rapid-fire exposures and comparing them, you can infer some more detail by looking at overall motion and fine transitions. This technique has been used on Pixel devices (at least, possibly others) and Google (former employer) calls it “super-res zoom”.
Like traditional noise-based oversampling. When you have an analogue signal going into an ADC with a known noise distribution filling it, you can use that noise to infer bit values beyond the resolution of that ADC.
Dude your videos are so, so good! The graphics you track down are super cool and informative, and the explanations have great flow and are really intuitive. Plus you pick interesting topics! Keep it up!
Putting a bunch of images together to get the effect of a big lensed DSLR camera works pretty good.
Good report.
This is one of the most brilliant videos on TH-cam. Congratulations! Supposing the zoom issue is resolved with the next iphone 15 or Samsung or whatever, one wonders what is left to improve in smartphones? Maybe microscopic images 100X? Who needs a telescope or microscope? Maybe you could have an image at night and zoom into a star in the sky and also into the cells of a leaf in the same image. After the image is saved…
Bonsoir, toujours des super vidéos. Quelle synthèse. Vos compétences sont immenses. Bravo
Great video. The quote at the end, cherry on top of the cake!
The reason why the background is blurred in your picture at 15:45 is not software but because the bigger sensor gets you shallower depth of field
Depth of field is determined by the focal length of the lens and its aperture it has nothing to do with sensor size.
The camera should take the most accurate photo possible. I'm aware that all engineers see differently, and that there are a lot of variables, but effects should be added after an accurate photo has been taken. I use Light Room and all kinds of other software, but require a raw, accurate, uncompressed photo to start with.
Perfecto!!! This is my first time on your channel. Your explanation and overall video was so good that I was forced to liked and subscribed. Thanks man
This was SUCH a good episode, especially the bit of thought philosophy near the end. How often we, as Humans, ask for or prefer something other than reality. I am quite sure this says something profound about our species and the environments we experience.
Its the one aspect of digital photography which I really dislike.....its synthetic, a maze of mathematical computation which approximates reality, in other words....fake.
Looks like "folded zoom" concept is very similar to those camera optics in the RF-5E recon jet. The light gets folded 90° thrice. Allowing the camera to zoom all the way to 50 nautical miles slant range while still managing to cram 3 different cameras + stabilizing systems in the nose.
@27_ECE_Gokul V Slant range is basically the direct line-of-sight range to target (hypothenuse of triangle). We specify "slant" because a plane flies high up.. and not at the same horizontal level to object (like base of triangle).
Yes it can clearly see human-sized objects >50NM away despite being all-mechanical and using film. Much of the capability can be attributed to the stabilizer system. Targeting pods today cannot match that due to pylon vibration and pod design.
It was the best recon camera the US ever produced until 1990s SYERS 2 came on upgraded U-2s and satellites. Ironically it was never used by the US.. first customers were Malaysia and Saudi Arabia. Malaysia later upgraded to even more capable digital camera.
Very nice. Now do panoramas! I know they can line up overlapping parts of neighboring images, but I always wondered how much they use accelerometer data to help them guess how far you've moved/tilted between the images.
@15:47 the new larger sensor of the iPhone’s standard (1x) camera is actually big enough for fully natural bokeh to occur with close-up subjects. This is most likely what you’ve noticed creeping.
Great video. It's a shame smartphones can't operate external flashes and generally lack the easy to control manual mode that DSLR cameras have. So convenient though. Glad to have both.
Do not surprise if you soon need to pay extra (subscription) for using cameras (software/bios/firmware). Compact cameras with alternative bios/firmware unlock great capabilities (CHDK for canon for e.g.). Same apply to all digital camera, but "dumding down", control/removal of ownership making tech crappier and crappier.
Sony phones have excellent pro mode. But I doubt anyone buys Sony phones.
Really great video. I forgot to watch it during Patreon Early Access. But it's a really nice deep dive!!
I think i recognize the scenery in those nightime shots in 12:24 and 12:30!! I think its in Taipei for the former and Sunset scene in Gaomei wetlands for the latter. Great pics!
I remember when I first saw a picture from a camera on a flip phone in 2003. It was absolute garbage and I literally laughed at how bad it was. The improvements over the intervening 2 decades - the CCD to CMOS switch, back-illuminated sensors, pixel crosstalk suppression, quantum-dot based sensors - have been incredible. There's still enormous room for improvement yet, only a couple percent of the area of the back of a modern cameraphone is actually collecting photons...
Deformable lenses.
What I don’t understand is why can’t they make square sensors? That would allow taking landscape photos while holding the phone in more convenient portrait orientation.
We cloud mentioned laser auto focus functions I think first introduced LG G3. Apple's lidar tech. Some sensor sets even have special color sensor to make sure right color and white balance.
I finally got a stand alone camera. So glad I don't have to upgrade every year and holding a camara feels so much better than something made for talking.
15:54
that the iPhone "blurr" the background has nothing to do with portrait mode but literally optics... The iPhone 13 Pro sensor is much bigger than previous iPhones and that's why the natural bokeh is much more noticeable. This is just a physical aspect of the optics and is not some software that Apple applies to the picture. That's also why it's only on close-ups and nowhere else.
15:43 - what you're seeing there is actual bokeh. With smartphone sensors becoming larger and larger, and with the optics still being tiny, the depth of focus with high-end smartphone cameras has become extremely shallow. This is actually a problem with a lot of phones because it's getting to where the depth of field is too shallow for the entire image to be in focus. We've been starting to see workarounds to this problem being adopted such as shifting over to the ultra-wide lens when shooting very close subjects to ensure that the whole image is in focus (with the tradeoff of lower image quality). My S23 Ultra has this feature (Samsung calls it "focus enhancer"), and I believe the Pixel 7 Pro also does.
Love your videos. Is that Halcyon at 15:20 ish?
Suggest you deep-dive camera module packaging, the integration of optical mechanical and electronic components and fictions in a miniaturized package is a marvel of miniaturization and more than meets the eye.
I have four mirrorless Sony cameras. The last time I use one was about 4 years ago when I got my first pixel phone. My pixel 4 takes better pictures than any Sony I have. By better I mean more usable better color better shadow details they just look better.
I don't understand why Sony or Canon, etc, doesn't put the same processing in one of their mirrorless cameras. Their sales are going down down down because practically speaking they're not as good as most people's phone.
I refuse to spend 3 to 10 minutes in Lightroom to get a couple of photos looking as good as from my phone.
I know there's a huge, actually very small, group of photographers who want raw only and want to spend their lives photo processing. But to heck with them.
Watch this video, Sony.
Doesn’t the a7s series have leading low
Light tech? Or am I too far behind now
In the best way possible your mom is right about your videos putting people to sleep, Whenever you post I make sure to watch your videos as a way to wind down at the end of a long day. Thank you
Very well explained! I didn't learn much from this particular episode but I've learned a great deal from your others. Thank you.
Fantastic in-depth presentation.
Saying that a phone is better than a camera is like saying a moped is better than a car
your research and documentary are very very informative!! Thank you for making such videos
Imagine all of that engineering, and no one thought to put a sticker on the camera reading "No one gives a shit about your facebook pictures"
Great video. Does computational photography correct perspectives of different focal lengths? Due to the small size sensors, an 85mm focal length equivalent on the smartphone is not the same as 85mm on a DSLR. In other words, the smartphone crops a wider-angle portrait into a DSLR's 85mm portrait look-alike. Will the computer inside the smartphone correct the inherent shortcomings of a wide-angle portrait (flattened look, protruding nose, etc)?
If I lug around my MILC camera with a couple good lenses I almost always get great pics. The only shots I've been asked to print and frame were taken with my camera. But you're lugging around a couple grand of equipment for that experience. I've definitely left it at home because of the weight or worrying about keeping it somewhere while on a trip.
It really is amazing how good phone cameras are now, so I don't anguish much over leaving the big camera at home anymore. Ultimately the best camera is the one you have, and 99% of the time that's going to be a phone camera that'll do the job 90% as well.
I totally agree. Based on how my pics were being consumed (95% of the time on people's phones, with the very rare print), and the hassle of schlepping the lenses around, I sold my Sony a7r3 a couple of years ago. I did buy a buy an RX10M4 as a kind of compromise between full frame and a phone, but I rarely ended up using it. On a recent trip to Vietnam, i just took my Samsung S22 Ultra, and I'm perfectly happy with the results (though I'd really miss the 10x optical telephoto if I had an iPhone).
(All that said, I just read a review of the Canon R7, and I'm sorely tempted, especially since I still have a lot of EF mount lenses from my pre-Sony days!)
Been waiting for this one - I appreciate glass more and more because of this channel
All my photos taken on any digital device seems like a lie now. I abandoned the older way because I thought the future was digital photography.
Goes to show you, old school lenses are the only way to get actual pictures from a object or subject.
Great video and compilation of history. I literally was there feeling the revolution, but never explored the tech behind it. Interesting one!
Smartphone cameras have come a long way, just 10years before, except some handful devices camera quality was plain pathetic in most of the devices but now their achievement is in front of us. However still i do feel that computational photography can't beat the laws of physics, they are still very profound. AI /ML creates very unrealistic looking images , oversharpen unnecessarily and colour accuracy is gone for a toss. Most frustrating though is loss of depth in the images, they looks flat and washed out which is something nokia lumia 808 was known for
Photos from my Nokia N73 and N95 still look absolutely gorgeous.
A lot of new information - thumbs up!
very well explained! thank you for that!
I'm confused, were you saying that core photonics was the first company to float the idea of putting two different cameras (zoom and wide) into a phone body? Or were they the first company to develop a practical zoom camera in a size that would fit a phone? I'm assuming Samsung bought them for their folded optical path IP more than anything else.
small correction: in 15:00 you talk about the fokus effect, and how smartphones cant have a natural one. While thats true for most, like my former honor 8 which had 2 cameras just for to mimic this effect, my now owned Xiaome Note 10 does actuelly have a natural fokus effect. a pretty strong one too, sometimes it sucks, but when used on porpuse its amazing. Its duo to the bigger sensor and lenses, someone told me, idk really
Thanks a lot for these explanations !!! I had asked myself this question a lot, as I have an SLR camera, and I was thinking that some pictures looked better taken with a smartphone...I suspected that computer power was involved!
2 cents: depth of field and Bokeh are two different concepts. Your comparison of portrait mode and the blurry background of the drink picture let me know that you are confused. But I am sure some quick online search would clarify them to you.
Not that long ago (seemingly) I remember thinking how cool it was to snap a pic with a Motorola Razor. 😂😂
Great video again! Love your work
That was a great video, thanks! I also found it very curious that you had a picture of my subway station in Sofia, Bulgaria, have you been there?
dude i was just trying to google how camera sensors actually work and trying to brainstorm how that would be. this was rreally interesting and well made. thanks for a great video! take my like and subscription!
Very nice channel. I have seen a few videos now - including the very well-researched "The Rise and Fall of ICL". Thank you. I have a feeling I'm going to binge watch your previous ones! :-) All the best, Rob in Switzerland
Very interesting analysis, thanks for this!
This really is an awesome video, thanks for your work!
Impressive video. and the message at the end was so true. People want to see what they want not what it is. ;)
15:43 there's no AI bokeh there. It's the normal depth of field since the large aperture and big sensor.
The bokeh blur on the macro photo is optical, not digital... When the object is close enough you can have a nice blurred background :)
Well done tutorial - Thank You!
I loved how you ended this video very well said and thought provoking, Many camera phones apply skin softening in portrait mode - crazy to think all our memories will have been doctored - it's like something from 1984.
By the way, your mom is 100% correct I usually fall asleep to one of your videos playing. It's always a video that I've seen before so I'm not ignoring you.
First time and first video I've watched on your channel. Really informative and technical but at an accessible level, so the last thing I felt when watching this video was sleepy!
To repeat what others have said, I'm a bit of an amateur photographer and a lot of photography is figuring out how to frame the shot just right that it only shows what you want it to. Making careful use of zoom and positioning while out taking photos, then in Lightroom or a lightroom if using film, messing with the raw photo. DSLRs don't have much computing power but computers do. I do the same thing with my photos as the phones do. Adjusting saturation and exposure and toying with denoising to get it just right.
That philosophical conclusion doe.
Food for thought indeed
i have discovered this channel far too late. great video!
Yet another excellent report ! thanks !
That is very illuminating. Thanks.
Thanks for the video!
My 13 pro max camera is really nuts. The pictures are just perfect.
All the cameras before were nice upgrades and did a great job, but you'd always find something to complain about. Since the XS the pictures often look quite artificial, the auto HDR didn't always do a great job, uneven exposure taking panoramic photos, video stabilization was often stuttery, iPhone 11 pro only main camera was useful in the dark, the exposure was taking quite long...
The iPhone 13 pro is just perfect, no complaints. It's insane how heavy the picture processing is yet it looks very natural and colors and exposure are just on point.
Not just the camera, everything is perfect about this phone, before that there always was something missing. Perfect screen (120hz, 1000nit), perfect battery, perfect camera, 5g and 5ghz hotspot, perfect speakers. Only usb-c missing, though lightning is great. My iPads usb-c port doesn't work properly anymore.
I'm still waiting for DSLR or mirorrless cameras to get automatic frame stacking like smartphones do, it makes a huge difference. I still want "raw" photos but with more detail. Just like when you stack images in DeepSkyStacker for example. Better dynamic range, better signal to noise ratio. It's too slow doing it manually with a DSLR plus a lot of big cameras don't take fast enough bursts. Today phones have usually better dynamic range than even full-frame cameras.
some cameras that use IBIS got high res modes. Pentax, Olympus.
Dual Camera system was first used by HTC on their One series. But not for zoom only portrait mode
I'm pretty sure Google's been doing the image stacking thing far before 2018, it's called HDR+ and I think was present on the 2013 Nexus 5. HDR+ is where the phone is constantly taking frames, before the button is pressed. Night sight is when it does this only when the button is pressed.
Interestingly, all of my photos, taken using dedicated interchangeable lens cameras, are commented on by people quite often, s “that doesn’t look true to the scene”, while it actually is… according to a single image, taken with a limited dynamic range sensor, the processing pipeline behind it and all baked in… Conversely, this is what we consider as “taking a picture”, while what people apparently desire are the results of “making a picture”, enter the smartphones with cameras, no post-editing of single- and multiple- photographs. This begs the question “does the difference described here - taking vs. making - really matter? Most would say NO, while most photographers and enthusiasts hold on to the original single image outlet of a dedicated camera. No matter, the image quality differences re still visible in a way, where I can easily spot images made with a phone, vs. those taken with a camera - conclusively, I’d never take any important image using a phone, too much of a compromise on quality and editing potential…
What I wonder is who makes the flat lenses for high resolution sensors. It”s easy to find ones for 2MP but high resolution wide angle lenses are quite bulky
iPhone 13's camera has a large enough physical aperture that the bokeh in close-up photos such as that glass is formed optically. It's not unusual to stop the aperture of a traditional camera's lens down to f/8 and smaller to get useful DOF in macro shots.