[smcv]

From your description, I understand that your team apply Pixel binning to Slow-mo mode. Do you have an intention to apply this technique to create a oversampling mode or night mode for photo? That would create better photo in lower resolutions.

Hi, that's how the photo viewfinder runs today.. this is why we are able to keep it running above 30fps and still get bright preview in low light at that exposure time. I don't think we plan to allow for photos with binning, but it's definitely an idea worth exploring as long as people would be happy with 4.5MP photos, and the associated artifacts.
In the current low-light mode, we apply temporal oversampling -- 4 frames are captured and composited to reduce noise and motion blur, rather than using a long exposure (since you cannot remove blur from a long exposure).

[smcv]

I have another question : does the phone camera sensors outputs unuseable pictures without heavy post-processing ?

You can see what un-postprocessed photos look like when you use manual exposure controls. This disables the low-light frame stacking. There will always be some image processing of course, such as de-noising and sharpening. This is just the reality with these small mobile camera sensors. If you look at a DSLR camera like the Nikon D4s, that may have 7.3um pixels, those pixels have 42x more area than the 1.12um pixels in Priv's camera, and so are 42x more sensitive to light.. that's almost 5.5 stops of additional sensitivity, and so in the same lighting conditions, that camera would have 5.5 stops less noise (the equivalent of ISO 50 vs ISO 2100!). This is why all mobile phones need to innovate on the image processing front, whereas on a DSLR, the raw output is a very nice starting point, and not as much noise reduction is needed.

[guizmox]

...and that's exactly why I don't understand why Sony tries to put the maximum pixels on their tiny sensors.
> If it's only for zooming options, a digital zoom is not a real zoom as it does not decrease the depth of field.
> If it's for commercial purposes, it's a bad idea

I would like to read your thoughts about a sensor from a phone-camera with only a few mpixels on it (bigger pixels & bigger space between them). Do you think it could drastically increase the quality/sensivity ?

[smcv]

...and that's exactly why I don't understand why Sony tries to put the maximum pixels on their tiny sensors.
> If it's only for zooming options, a digital zoom is not a real zoom as it does not decrease the depth of field.
> If it's for commercial purposes, it's a bad idea

I would like to read your thoughts about a sensor from a phone-camera with only a few mpixels on it (bigger pixels & bigger space between them). Do you think it could drastically increase the quality/sensivity ?

Everyone in the know is typically on the same page regarding digital zoom. It's more of a convenience feature, as it destroys quality. Optical zoom is a completely different story
Sony has some tiers of phone which are pushing very high megapixel counts, but I personally believe that this is due to the fact that Sony actually builds camera sensors and sells them to other manufacturers, so they need to showcase their product line.

Bigger pixels definitely improve quality. Take a look at the Nexus 6P reviews for example. They went to a 12-ish MP 1.55um pixel camera whereas a sensor like Sony's 21MP model has 1.12um pixels. While the Sony can resolve a bit higher detail due to pixel density, a 1.55um pixel has DOUBLE the light collecting capacity. This means half the noise in normal photos, and twice as bright photos in lower light. As well, it should consume less bandwidth and power due to pushing fewer pixels per second around.

I think you can take a look at camera specs from 2015 to 2016 and draw your own conclusions about the trends.

FYI, pixel binning effectively gives us 2.24um pixels on Priv, which means that preview (or 60fps video) can be 4x brighter for the same exposure time, or have 4x less noise.. this lets us run autofocus much faster because we can crank up the framerate, and focus accuracy is improved because there is less noise to interfere with the contrast calculations. Resolution suffers due to the geometry of the bayer mask though -- the 2.24um pixels are spread over a 4.48um area for reds and blues, greens are tighter.

[ajwan]

Would it be possible to use binning to increase sensor sensitivity (and to decrease noise) while at the same time reducing MP (an acceptable tradeoff to me anyways)? Could this be implemented via software in an update?

Thought I'd ask.

[Lawmen23]

Wow...I feel so much smarter now after reading this thread. I do have a question for SMCV though. It's apparent to me that the front facing camera only takes great pictures in ideal lighting. In the age of prevalent social media apps and the popularity of "selfies", what was the logic used behind choosing that sensor or the programming associated with it? Was it just a cheap sensor to reduce hardware costs, or can it be brought up to speed with additional programming? Thanks in advance. I'm still on Lollipop by the way(thanks Verizon), so if there was an improvement in the recent update I may not have it.

[smcv]

Would it be possible to use binning to increase sensor sensitivity (and to decrease noise) while at the same time reducing MP (an acceptable tradeoff to me anyways)? Could this be implemented via software in an update?

Possibly.

[smcv]

Wow...I feel so much smarter now after reading this thread. I do have a question for SMCV though. It's apparent to me that the front facing camera only takes great pictures in ideal lighting. In the age of prevalent social media apps and the popularity of "selfies", what was the logic used behind choosing that sensor or the programming associated with it? Was it just a cheap sensor to reduce hardware costs, or can it be brought up to speed with additional programming? Thanks in advance. I'm still on Lollipop by the way(thanks Verizon), so if there was an improvement in the recent update I may not have it.

The answer to the front camera question is two-fold: mechanical and schedule impacts. The design cycles for these phones are measured in double-digit months, and so changing course late in the game would have had consequences. Remember also, the iPhone only had a 1.2MP camera until a few weeks before Priv launched.
We developed the selfie pano feature as a software improvement to help get more performance out of the front camera (4MP or so).

[gizmo21]

Would it be possible to use binning to increase sensor sensitivity (and to decrease noise) while at the same time reducing MP (an acceptable tradeoff to me anyways)? Could this be implemented via software in an update?

Thought I'd ask.

Would love to see that in the manual mode menu (so standard user isn't irritated), or even as permanent option in Prefs like the video modes are set there.

I'm a proud user of a (already dated) compact Lumix LX-3 which was introduced with only an 10MP sensor which was against the trend at that time and i still chuckle when i take great lowlight pics where others struggle with flash on their 20MP compact cameras.

I wonder how the team will manage to implement all that special sensor specific stuff once you have to deal with, let's say perhaps TWO more european-city devices?

[maxivn]

I think it would be great if your team could implement pixel binning + EIS (electronic image stabilization) + OIS (optical image stabilization) into manual mode for better long exposure picture (as it only limits to 1/2s in the newest update). I see the new ASUS zen has combined EIS and OIS. Is it possible for implementation with Priv hardware?