3 ms·

are able to convincingly fake something like an 85mm f/1.4, If that were the case we have a good long time then, most likely; Really what you are talking about

by ska 7y ago

are able to convincingly fake something like an 85mm f/1.4, If that were the case we have a good long time then, most likely; Really what you are talking about is how you can't beat the physics of large sensors and lenses; for now 'faking it' isn't anywhere close, but also the techniques being pursued aren't likely to get there. I suspect in-camera processing though is a bit of a dead end for pro and prosumer cameras because your laptop/desktop is going to be able to do a much better job, and your camera just isn't going to have a good workflow for post-processing and organizing photos. For phone camera use it makes more sense, because people are rarely doing much more than delete these two, crop this one a little and send it to mom. Don't get me wrong, phone cameras are having a ton of effort put into them and have way more resources than conventional cameras - but the effort is almost entirely being put into making something small and light work pretty well under most circumstances, not into making something great. What's more important is that most people don't care for most applications. More people use to buy "good" cameras because their other options were poor or nonexistent. A top of the line current phone camera may not be great, but it's good enough for most people. I don't know what this means. It could well kill off the "prosumer" market and leave professional cameras an expensive niche product. It just likely won't be because phone cameras got as good at creating pictures.

falcrist 7y ago
Yea. You can't beat the physics of big optics... but there ARE ways to get around some of the limitations of smaller optics. You can fake depth of field, use image stabilization and image stacking to gather more light with a smaller aperture, utilize "super resolution" and/or neural network techniques to get or fake more detail, etc. I'd love to see some of the awesome work being put into phones carry over into full-frame cameras before the whole pro-sumer market and a bunch of the pro market dissolve.
- ska 7y ago
  Image stabilization and stacking both started in SLR's before phone cameras were doing it. The best image stabilization is still opto-mechanical, of course. Nobody fakes depth of field that well, and super resolution (the algorithmic kind, not the real kind) doesn't work that well either. We've been doing research in those areas for 25+ years now, and while some real progress has been made it's not magic. Phone cameras are doing a lot of heroic stuff to get around the limitations of their optics, and it results in "hey, that's not bad". Nothing wrong with that! But I don't see why you would want to see that stuff in a digital SLR, say, though - if you take your photography seriously enough to lug that thing around, you're doing post processing on a computer anyway, and you can do all that stuff better there (at least in theory). Now if you were asking why photo processing software on your desktop wasn't absorbing or improving on everything in the phones, that would make more sense to me.
  - falcrist 7y ago
    Well first of all, there are things that can be done in camera better than they can be done in post in any reasonable amount of time. Image stacking is one of those, since alignment even with the best software is spotty, and cameras could include fairly robust accelerometers to do a bunch of the work in the background. There's also currently no depth info in DSLR images, so image processing involving depth of field isn't going to work unless you do the work by hand, which is quite laborious. In addition, it would be very nice if HDR pictures could be taken by my full-frame camera. I don't necessarily want or need to deal with bracketing in lightroom just to get an extra couple stops of dynamic range. You're right that the best image stabilization is opto-mechanical, but it would be very nice to have access to digital image stabilization as well as in-body and in-lens stabilization. Again, a decent accelerometer can go a long way. Honestly, most of what I'd like to see has involved image stacking in one form or another. Digital ND filters are the same idea. And don't get me started on ergonomics and usability. I definitely picked the worst camera brand in terms of user interface (sony), but all the DSLRs and mirrorless cameras I've touched are noticably lacking in that department. I don't necessarily need a giant phone screen like my S9+, but I would expect a device costing $2000+ to at least have a fairly high resolution OLED panel on the back of it with solid touch controls. Meanwhile touch controls are distinctly lacking in any camera I've tried. Software is important too!
    - sudosysgen 7y ago
      To be fair, Sony cameras have high resolution OLED panels, in the viewfinder.
    - cozzyd 7y ago
      I got a Canon Rebel SL3 recently (APS-C, not full-frame, but still...) and it can do HDR on-camera. I wish it would save each individual exposure too instead of just the composite though. It also has touch controls on the display and a REST API for remote control via wifi. And this is not that expensive a camera.
    - numpad0 7y ago
      Sounds like another direction to expand RAW formats. Accelerometer data is small, burst shots wouldn’t be that hard to take diff and store along. Then you could hypothetically feed it to Photoshop 2022 and run a 3D reconstruction.
scottlamb 7y ago
> I suspect in-camera processing though is a bit of a dead end for pro and prosumer cameras because your laptop/desktop is going to be able to do a much better job, and your camera just isn't going to have a good workflow for post-processing and organizing photos. Something like Night Sight as a post-processing method might not be practical. You need many short exposures instead of one long exposure. So basically, you've got full-resolution raw video to deal with instead of a single shot. More efficient to process that before storing it and transferring it.
- nucleardog 7y ago
  Based off of how cameras are currently used (at least from my limited sample size), I don't think that would really matter. People regularly shoot in various "raw" formats which are significantly larger in size so that they have more flexibility in how to post-process. They capture a whole bunch of extra data, move it around, and generally store it indefinitely knowing they will throw a lot of it away in the final photo so that they can make the decision what to throw away later while they're looking at a 27" colour-corrected screen instead of a small camera screen. People don't do HDR on-camera, they generally shoot multiple exposures and adjust in software after to get the effect and balance they're looking for which is basically Night Sight to a lesser degree. A bit of extra disk space and transfer time is small peanuts and already fits the workflow most photographers are used to.

by ska 7y ago

falcrist 7y ago

Yea. You can't beat the physics of big optics... but there ARE ways to get around some of the limitations of smaller optics. You can fake depth of field, use image stabilization and image stacking to gather more light with a smaller aperture, utilize "super resolution" and/or neural network techniques to get or fake more detail, etc. I'd love to see some of the awesome work being put into phones carry over into full-frame cameras before the whole pro-sumer market and a bunch of the pro market dissolve.

ska 7y ago
Image stabilization and stacking both started in SLR's before phone cameras were doing it. The best image stabilization is still opto-mechanical, of course. Nobody fakes depth of field that well, and super resolution (the algorithmic kind, not the real kind) doesn't work that well either. We've been doing research in those areas for 25+ years now, and while some real progress has been made it's not magic. Phone cameras are doing a lot of heroic stuff to get around the limitations of their optics, and it results in "hey, that's not bad". Nothing wrong with that! But I don't see why you would want to see that stuff in a digital SLR, say, though - if you take your photography seriously enough to lug that thing around, you're doing post processing on a computer anyway, and you can do all that stuff better there (at least in theory). Now if you were asking why photo processing software on your desktop wasn't absorbing or improving on everything in the phones, that would make more sense to me.
- falcrist 7y ago
  Well first of all, there are things that can be done in camera better than they can be done in post in any reasonable amount of time. Image stacking is one of those, since alignment even with the best software is spotty, and cameras could include fairly robust accelerometers to do a bunch of the work in the background. There's also currently no depth info in DSLR images, so image processing involving depth of field isn't going to work unless you do the work by hand, which is quite laborious. In addition, it would be very nice if HDR pictures could be taken by my full-frame camera. I don't necessarily want or need to deal with bracketing in lightroom just to get an extra couple stops of dynamic range. You're right that the best image stabilization is opto-mechanical, but it would be very nice to have access to digital image stabilization as well as in-body and in-lens stabilization. Again, a decent accelerometer can go a long way. Honestly, most of what I'd like to see has involved image stacking in one form or another. Digital ND filters are the same idea. And don't get me started on ergonomics and usability. I definitely picked the worst camera brand in terms of user interface (sony), but all the DSLRs and mirrorless cameras I've touched are noticably lacking in that department. I don't necessarily need a giant phone screen like my S9+, but I would expect a device costing $2000+ to at least have a fairly high resolution OLED panel on the back of it with solid touch controls. Meanwhile touch controls are distinctly lacking in any camera I've tried. Software is important too!
  - sudosysgen 7y ago
    To be fair, Sony cameras have high resolution OLED panels, in the viewfinder.
  - cozzyd 7y ago
    I got a Canon Rebel SL3 recently (APS-C, not full-frame, but still...) and it can do HDR on-camera. I wish it would save each individual exposure too instead of just the composite though. It also has touch controls on the display and a REST API for remote control via wifi. And this is not that expensive a camera.
  - numpad0 7y ago
    Sounds like another direction to expand RAW formats. Accelerometer data is small, burst shots wouldn’t be that hard to take diff and store along. Then you could hypothetically feed it to Photoshop 2022 and run a 3D reconstruction.

scottlamb 7y ago

> I suspect in-camera processing though is a bit of a dead end for pro and prosumer cameras because your laptop/desktop is going to be able to do a much better job, and your camera just isn't going to have a good workflow for post-processing and organizing photos. Something like Night Sight as a post-processing method might not be practical. You need many short exposures instead of one long exposure. So basically, you've got full-resolution raw video to deal with instead of a single shot. More efficient to process that before storing it and transferring it.

nucleardog 7y ago
Based off of how cameras are currently used (at least from my limited sample size), I don't think that would really matter. People regularly shoot in various "raw" formats which are significantly larger in size so that they have more flexibility in how to post-process. They capture a whole bunch of extra data, move it around, and generally store it indefinitely knowing they will throw a lot of it away in the final photo so that they can make the decision what to throw away later while they're looking at a 27" colour-corrected screen instead of a small camera screen. People don't do HDR on-camera, they generally shoot multiple exposures and adjust in software after to get the effect and balance they're looking for which is basically Night Sight to a lesser degree. A bit of extra disk space and transfer time is small peanuts and already fits the workflow most photographers are used to.