Can a Pixel phone outresolve a G9ii with a Pro lens?

[UrbaneHobbit]

I believe all modern smartphones are doing some form of multi-image capture and fusion, as described in this article from a few years ago:

You're already using computational photography, but that doesn't mean you're not a photographer

In our second 'Smarter Image' column, we'll take a look at the core features that make modern smartphone photography possible.

www.popphoto.com

I suspect that this may be why the HHHR output from the G9 II compares better than a single exposure from that camera.

I hope to see MFT adapt more computational photography technologies to stay ahead of smartphones, so that we can take advantage of some of the higher-end lenses in the system. Whether this is possible with current market trends for chips, readout speeds vs. tiny smartphone sensors, and heat challenges involved in running these image processing pipelines, I don’t know.

 

[Will Rose]

Just the fact that this conversation can be reasonably sustained and debated shows how far phone cameras have come.

The comparison so far is, indeed, interesting.

However, and I don’t think this was the OP's intent, I think the subject in this test is quite biased in favor of the phone.

The subject, although a real window, is not a real-word subject: it’s more like a painting or animation. It has large swaths of color punctuated by sharp line and some details. It is not highly textured or highly detailed in low contrast areas, like many natural scenes are.

It would be more telling to see how the competitors fare with more traditional subjects:

1. A face that fills the frame. Let’s see how well the low-contrast detail of skin is captured.

2. Grass. An outdoor scene that has a bold main subject, but also shows high detail on in-focus grass and leaves. Older phones tended to smear grass into mush.

3. Fabric. The old Imaging Resource test was excellent. Can the sensor resolve, and the processing preserve, the low-contrast detail in the weave and pattern of red and orange fabrics?

And what in the world happens when the ISO goes up?

The above is a test that would be well worth seeing.

But this comparison, because of the content, cannot be conclusive.

The phone was able to show enough to get in the door. Now let the real competition begin.

 

[Owlsfan]

I agree tbh. It wasn't the best subject to test. Maybe the results will be the same though but my pixel 7 wasn't particularly amazing.

My current phone is really good at night .But even my cheap lenses show more fine detail in good light.

 

[HicHic]

I agree tbh. It wasn't the best subject to test.

What's the best subject to test?

 

[Will Rose]

I agree tbh. It wasn't the best subject to test.

What's the best subject to test?

I provided three:

1. A face that fills the frame. Let’s see how well the low-contrast detail of skin is captured.

2. Grass. An outdoor scene that has a bold main subject, but also shows high detail on in-focus grass and leaves. Older phones tended to smear grass into mush.

3. Fabric. The old Imaging Resource test was excellent. Can the sensor resolve, and the processing preserve, the low-contrast detail in the weave and pattern of red and orange fabrics?

 

[HicHic]

I agree tbh. It wasn't the best subject to test.

What's the best subject to test?

I provided three:

1. A face that fills the frame. Let’s see how well the low-contrast detail of skin is captured.

2. Grass. An outdoor scene that has a bold main subject, but also shows high detail on in-focus grass and leaves. Older phones tended to smear grass into mush.

3. Fabric. The old Imaging Resource test was excellent. Can the sensor resolve, and the processing preserve, the low-contrast detail in the weave and pattern of red and orange fabrics?

1 - not a still subject, changing variable - someone's gunna complain.

2 - outdoors, haze, atmosphere changing, different lighting, someone's gunna complain.

3 - harder to tell if/when the Pixel phone is creating fake textures to appear more detailed - someone's gunna complain.

 

[ahaslett]

I agree tbh. It wasn't the best subject to test.

What's the best subject to test?

I provided three:

1. A face that fills the frame. Let’s see how well the low-contrast detail of skin is captured.

2. Grass. An outdoor scene that has a bold main subject, but also shows high detail on in-focus grass and leaves. Older phones tended to smear grass into mush.

3. Fabric. The old Imaging Resource test was excellent. Can the sensor resolve, and the processing preserve, the low-contrast detail in the weave and pattern of red and orange fabrics?

Highlights on a brown bottle that are not blown out are a pretty harsh test of tonal gradation. Reflections of clouds in shop windows also.

Grass is pretty fractal - hard to massage grass and distant foliage without artefacts.

A

 

[cba_melbourne]

No it can't. Details get lost. It comes close though. This is a really huge zoom factor:

...I think the photo on the right is the Pixel photo.

And you are right. I mixed them up. Very embarrassing....

Question now is, it the Pixel that good, or is the G9ii that bad? Is the extra detail shown by the pixel real, or did it invent it?

I'm very impressed with the Pixel's results, and it isn't even the best phone out there for camera performance. The Vivo x200 sits at the top, and many other Chinese phones fill the performance gap between that and the Pixel.

The Pixel is just used as an example because it's the phone I have. It certainly isn't the best performing smartphone camera. The G9ii on the other hand, there's no dispute that it's the best M43 camera for resolving power.

I'm quite dumbfounded by how the Pixel is able to achieve such a result. The price I got for it new is pretty much the same cost as a new 12-45/4 Pro. Are the lenses in these phones that good? To imagine I had trouble finding a good copy of the 12-45 lens...with so much more glass, size and weight, only to be outperformed by some tiny lens in this phone. I don't see any decentering issues with my phone's images either. Is it easier to produce properly aligned lenses for phones? I just don't get how a package this small is outresolving my G9ii with a Pro lens.

I'm sure these phones do some degree of "guessing" to create details at the pixel level, but I honestly expected a ton of weird artifacts and errors. There are super tiny details in the scene where the phone couldn't just have guessed right without the sensor+lens combo collecting genuinely sharp and high resolution raw data to work off of. For instance, the "diagonal lines below the hand" section that I've highlighted in the previous posts, you can tell that the single shot G9ii is only resolving a mushy patch. One would think that, whatever the Pixel's raw data sees, it can't possibly be better than that. Yet, somehow, it correctly resolves the diagonal lines. The source data must be of genuinely high enough quality in order for it to get that result - that part I'm most impressed with.

I don't know anything about these phone sensors, but if it's anything like the OM-1's supposed "high resolution 80MP binned to 20MP", then maybe there is a lot of potential to be unlocked. Because my Pixel has the option to either shoot in 12MP or 50MP.

The way I look at it (correct me if I am wrong), in the binned 12MP mode you get real raw files like in a classical camera that you can fully work with in PP. You can't have background blur and bokeh emulation etc in that mode.

Whereas in the 50MP jpg mode you have to accept or put up with fake details generated in-phone by software.

Whatever the Pixel's 50MP actually resolves, it's certainly higher than 25MP.

 

[JakeJY]

No it can't. Details get lost. It comes close though. This is a really huge zoom factor:

...I think the photo on the right is the Pixel photo.

And you are right. I mixed them up. Very embarrassing....

Question now is, it the Pixel that good, or is the G9ii that bad? Is the extra detail shown by the pixel real, or did it invent it?

I'm very impressed with the Pixel's results, and it isn't even the best phone out there for camera performance. The Vivo x200 sits at the top, and many other Chinese phones fill the performance gap between that and the Pixel.

The Pixel is just used as an example because it's the phone I have. It certainly isn't the best performing smartphone camera. The G9ii on the other hand, there's no dispute that it's the best M43 camera for resolving power.

I'm quite dumbfounded by how the Pixel is able to achieve such a result. The price I got for it new is pretty much the same cost as a new 12-45/4 Pro. Are the lenses in these phones that good? To imagine I had trouble finding a good copy of the 12-45 lens...with so much more glass, size and weight, only to be outperformed by some tiny lens in this phone. I don't see any decentering issues with my phone's images either. Is it easier to produce properly aligned lenses for phones? I just don't get how a package this small is outresolving my G9ii with a Pro lens.

I'm sure these phones do some degree of "guessing" to create details at the pixel level, but I honestly expected a ton of weird artifacts and errors. There are super tiny details in the scene where the phone couldn't just have guessed right without the sensor+lens combo collecting genuinely sharp and high resolution raw data to work off of. For instance, the "diagonal lines below the hand" section that I've highlighted in the previous posts, you can tell that the single shot G9ii is only resolving a mushy patch. One would think that, whatever the Pixel's raw data sees, it can't possibly be better than that. Yet, somehow, it correctly resolves the diagonal lines. The source data must be of genuinely high enough quality in order for it to get that result - that part I'm most impressed with.

I don't know anything about these phone sensors, but if it's anything like the OM-1's supposed "high resolution 80MP binned to 20MP", then maybe there is a lot of potential to be unlocked. Because my Pixel has the option to either shoot in 12MP or 50MP.

The way I look at it (correct me if I am wrong), in the binned 12MP mode you get real raw files like in a classical camera that you can fully work with in PP. You can't have background blur and bokeh emulation etc in that mode.

Whereas in the 50MP jpg mode you have to accept or put up with fake details generated in-phone by software.

Whatever the Pixel's 50MP actually resolves, it's certainly higher than 25MP.

It's not completely fake, it's actually really 50mp of data, so does provide some real resolution advantage over a 25MP camera. The issue is that due to the Quad Bayer color filter, the way it is demosaiced may result in some artifacts that isn't seen in normal Bayer cameras (although Bayer has its own artifacts like false colors). When it is binned in 12MP mode however it's the same as a normal Bayer cameras so looks more "natural".

 

[cba_melbourne]

then the Pixel 9 should retain the same details, and perhaps even show them even more clearly.

Right, that shows the limits of 25MP vs 50MP, but that doesn't change the point I made, there are other areas (which I pointed out explicitly) where the Pixel 9 is clearly missing details that both G9 ii examples capture. So the results are highly inconsistent, likely due to the test method (unless the Pixel is doing some processing and making up details in some areas).

You can cherry pick any part of the image you like. Some will show the HHHR to be better, some will show the Pixel to be better. But if you're comparing single shot performance vs Pixel, the Pixel outresolves the single shot G9ii in the majority of the parts of the image.

Is that because the Pixel always does multiple shots?

Is the pixel combining the information from several focal length cameras/sensors into one picture (it is assumed that some phones do, it's called deep fusion). The phone makers do not talk about those things, they are tightly guarded secrets as it gives them an edge over the competition. And huge R&D money goes into these tricks. Like if you zoom with a phone that has 3 fixed FL cameras, they merge the information from the different cameras to make it look like it was a continuous optical zoom. It's not just a pure cropping digital zoom as we know it from our cameras, it is far better than that.

 

[Will Rose]

1 - not a still subject, changing variable - someone's gunna complain.

2 - outdoors, haze, atmosphere changing, different lighting, someone's gunna complain.

3 - harder to tell if/when the Pixel phone is creating fake textures to appear more detailed - someone's gunna complain.

Come on, don’t be silly.

This is the MFT forum, someone’s gunna complain about EVERYTHING!

You just have to plow ahead anyway…

 

[cba_melbourne]

I agree tbh. It wasn't the best subject to test.

What's the best subject to test?

I provided three:

1. A face that fills the frame. Let’s see how well the low-contrast detail of skin is captured.

2. Grass. An outdoor scene that has a bold main subject, but also shows high detail on in-focus grass and leaves. Older phones tended to smear grass into mush.

3. Fabric. The old Imaging Resource test was excellent. Can the sensor resolve, and the processing preserve, the low-contrast detail in the weave and pattern of red and orange fabrics?

1 - not a still subject, changing variable - someone's gunna complain.

2 - outdoors, haze, atmosphere changing, different lighting, someone's gunna complain.

3 - harder to tell if/when the Pixel phone is creating fake textures to appear more detailed - someone's gunna complain.

Fabric but on a tilted plane. Then you see what the phone does with the out of focus areas to emulate bokeh. Today's phones still use one single plane as demarcation, what is further away gets the bokeh treatment, what is close does not. That creates many strange looking transitions. Maybe go back to Phillipreeve's X200 review to see the samples of such effects, like the window of this scooter that shows how surreal the lack of smooth bokeh transition can look in certain situations. You definitely want to use that phone in the 12.5MP binned mode for this picture, as then it will not do fake bokeh:



https://phillipreeve.net/blog/vivo-x200-ultra-the-death-of-the-compact-camera/

 

[cba_melbourne]

No it can't. Details get lost. It comes close though. This is a really huge zoom factor:

...I think the photo on the right is the Pixel photo.

And you are right. I mixed them up. Very embarrassing....

Question now is, it the Pixel that good, or is the G9ii that bad? Is the extra detail shown by the pixel real, or did it invent it?

I'm very impressed with the Pixel's results, and it isn't even the best phone out there for camera performance. The Vivo x200 sits at the top, and many other Chinese phones fill the performance gap between that and the Pixel.

The Pixel is just used as an example because it's the phone I have. It certainly isn't the best performing smartphone camera. The G9ii on the other hand, there's no dispute that it's the best M43 camera for resolving power.

I'm quite dumbfounded by how the Pixel is able to achieve such a result. The price I got for it new is pretty much the same cost as a new 12-45/4 Pro. Are the lenses in these phones that good? To imagine I had trouble finding a good copy of the 12-45 lens...with so much more glass, size and weight, only to be outperformed by some tiny lens in this phone. I don't see any decentering issues with my phone's images either. Is it easier to produce properly aligned lenses for phones? I just don't get how a package this small is outresolving my G9ii with a Pro lens.

I'm sure these phones do some degree of "guessing" to create details at the pixel level, but I honestly expected a ton of weird artifacts and errors. There are super tiny details in the scene where the phone couldn't just have guessed right without the sensor+lens combo collecting genuinely sharp and high resolution raw data to work off of. For instance, the "diagonal lines below the hand" section that I've highlighted in the previous posts, you can tell that the single shot G9ii is only resolving a mushy patch. One would think that, whatever the Pixel's raw data sees, it can't possibly be better than that. Yet, somehow, it correctly resolves the diagonal lines. The source data must be of genuinely high enough quality in order for it to get that result - that part I'm most impressed with.

I don't know anything about these phone sensors, but if it's anything like the OM-1's supposed "high resolution 80MP binned to 20MP", then maybe there is a lot of potential to be unlocked. Because my Pixel has the option to either shoot in 12MP or 50MP.

The way I look at it (correct me if I am wrong), in the binned 12MP mode you get real raw files like in a classical camera that you can fully work with in PP. You can't have background blur and bokeh emulation etc in that mode.

Whereas in the 50MP jpg mode you have to accept or put up with fake details generated in-phone by software.

Whatever the Pixel's 50MP actually resolves, it's certainly higher than 25MP.

It's not completely fake, it's actually really 50mp of data, so does provide some real resolution advantage over a 25MP camera. The issue is that due to the Quad Bayer color filter, the way it is demosaiced may result in some artifacts that isn't seen in normal Bayer cameras (although Bayer has its own artifacts like false colors). When it is binned in 12MP mode however it's the same as a normal Bayer cameras so looks more "natural".

It's more complex than that. Many phones take several pictures and stack them, and they take those stacked pictures from the several different FL cameras/sensors in the phone and very intelligently fuse them together, to generate your final picture. That is a lot of data and processing, so the result is a compressed jpg to do all these computations in real time.

I think much of this sorcery and compressing is left out when you select the 12.5MP binned mode (or pro-raw mode, depends on phone of course they are all subtly different and use different terms, but I think they all have a mode for "more demanding" photographers that need consistent results independent of subject).

I think 12.5MP is the resolution of today's flagship phones that you can consistently count on, within the shooting envelope of a phone, at the native focal lengths of the phones lenses, without artifacts or varying resolution across the frame, and with an editable raw output file.

Anything above these 12.5MP depends on the subject and many many very complex factors outside your control. It may be incredibly outstanding in some pictures, and disappointing in others. On average, it will make casual users very very happy. But photographers used to repeatable and consistent results with classical digital cameras like m43, may not be that happy.

 

[Tom Caldwell]

I used to go out with a regular walking group just a few years ago and possibly the phone cameras were not as powerful as they are today.

However I was told regularly that the images produced on their phones were very good with glances at what they looked like on screen. Indeed they usually did. So here was me, lumping my camera kit.

But when the conditions varied, less light or telephoto the phone cameras were put away for another day. Whilst my camera gear soldiered on quite magnificently.

Seems a bit like horses for courses to me.

No doubt the camera phones captured exactly what their owners needed - so why the fuss and bother when "real" mount system camera gear was serving a greater purpose?

The smaller-sensor mount systems were very popular but effectively failed not because of the sensor size domino but because there was a limited range of lenses on offer.

M4/3 poses a much larger domino because it is a much more mature and established system with a huge number of lenses of almost every size, type, and cost. As already noted it is more likely to be crushed by the craze for perfection as personified in the form of larger sensors.

But just how perfect do we need?

 

[cba_melbourne]

I used to go out with a regular walking group just a few years ago and possibly the phone cameras were not as powerful as they are today.

However I was told regularly that the images produced on their phones were very good with glances at what they looked like on screen. Indeed they usually did. So here was me, lumping my camera kit.

But when the conditions varied, less light or telephoto the phone cameras were put away for another day. Whilst my camera gear soldiered on quite magnificently.

Seems a bit like horses for courses to me.

No doubt the camera phones captured exactly what their owners needed - so why the fuss and bother when "real" mount system camera gear was serving a greater purpose?

The smaller-sensor mount systems were very popular but effectively failed not because of the sensor size domino but because there was a limited range of lenses on offer.

M4/3 poses a much larger domino because it is a much more mature and established system with a huge number of lenses of almost every size, type, and cost.

Of course m43 lenses are much larger and heavier than phone lenses. But their cost is a major factor too.

The Vivo X200 contains 3 main prime lenses:

14/6.9 FF equivalent

35/5.9 FF equivalent

85/8.5 FF equivalent

(a good and useful FL trifecta by the way)

Yes these are not that bright in aperture. But consider the phone takes a stack of pictures from each lens, and fuses these stacks of different focal length into one final picture. The stacking suddenly makes the lenses a lot brighter and also improves resolution and DR. And the fusing provides a seamless zoom experience from 3 prime lenses, and also improves brightness and resolution. Phones have indeed become quite good in low light.

These three lenses may together well represent US$100 or 150 of the US$1,000 price of the phone. And yes, if you upgrade that phone after 4 years these lenses are lost (or handed down with the phone to a kid). That is $38 per year for the 3 lenses. They are essentially disposable. And in 4 years there will be better phone lenses and you will be happy to upgrade instead of re-using the old obsolete ones.

Now consider three similar IQ m43 lenses of say 7, 17 and 42.5mm FL, what do these cost you to buy? What is their cost of ownership per year if each lasts say 10 years?

As already noted it is more likely to be crushed by the craze for perfection as personified in the form of larger sensors.

But just how perfect do we need?

 

[James Stirling]

I used to go out with a regular walking group just a few years ago and possibly the phone cameras were not as powerful as they are today.

However I was told regularly that the images produced on their phones were very good with glances at what they looked like on screen. Indeed they usually did. So here was me, lumping my camera kit.

But when the conditions varied, less light or telephoto the phone cameras were put away for another day. Whilst my camera gear soldiered on quite magnificently.

Seems a bit like horses for courses to me.

No doubt the camera phones captured exactly what their owners needed - so why the fuss and bother when "real" mount system camera gear was serving a greater purpose?

The smaller-sensor mount systems were very popular but effectively failed not because of the sensor size domino but because there was a limited range of lenses on offer.

M4/3 poses a much larger domino because it is a much more mature and established system with a huge number of lenses of almost every size, type, and cost.

Of course m43 lenses are much larger and heavier than phone lenses. But their cost is a major factor too.

The Vivo X200 contains 3 main prime lenses:

14/6.9 FF equivalent

35/5.9 FF equivalent

85/8.5 FF equivalent

(a good and useful FL trifecta by the way)

Yes these are not that bright in aperture. But consider the phone takes a stack of pictures from each lens, and fuses these stacks of different focal length into one final picture. The stacking suddenly makes the lenses a lot brighter and also improves resolution and DR. And the fusing provides a seamless zoom experience from 3 prime lenses, and also improves brightness and resolution. Phones have indeed become quite good in low light.

These three lenses may together well represent US$100 or 150 of the US$1,000 price of the phone. And yes, if you upgrade that phone after 4 years these lenses are lost (or handed down with the phone to a kid). That is $38 per year for the 3 lenses. They are essentially disposable. And in 4 years there will be better phone lenses and you will be happy to upgrade instead of re-using the old obsolete ones.

Now consider three similar IQ m43 lenses of say 7, 17 and 42.5mm FL, what do these cost you to buy? What is their cost of ownership per year if each lasts say 10 years?

As already noted it is more likely to be crushed by the craze for perfection as personified in the form of larger sensors.

But just how perfect do we need?

My brother has the xiaomi 15 ultra and the picture quality is very good a level beyond my Samsung S23 ultra and my daughters Iphone 16 pro max. He is not a keen photographer by any stretch but these phone cameras are getting better all the time. Making getting good results easy. No phone will ever achieve the flexibility of an ILC . But for what the majority of folk want it covers a lot of bases . The phone market is so big that they have massive resources for ever advancing products

My brother posts to our family Flickr account but nearly all his shots are of family and friends . Digging through his shots he has a few examples that don't include people





























--
Jim Stirling:
"To argue with a person who has renounced the use of reason, is like administering medicine to the dead." - Thomas Paine
Feel free to tinker with any photos I post

 

[HicHic]

1 - not a still subject, changing variable - someone's gunna complain.

2 - outdoors, haze, atmosphere changing, different lighting, someone's gunna complain.

3 - harder to tell if/when the Pixel phone is creating fake textures to appear more detailed - someone's gunna complain.

Come on, don’t be silly.

This is the MFT forum, someone’s gunna complain about EVERYTHING!

You just have to plow ahead anyway…

Come on, don't be silly.

I am doing exactly that, and testing the way I like.

 

[HicHic]

Fabric but on a tilted plane. Then you see what the phone does with the out of focus areas to emulate bokeh.

Not interested. I just want to analyze resolving power for the scope of this thread. Bokeh rendering for phones is not something that I'm interested to test.

 

[cba_melbourne]

Fabric but on a tilted plane. Then you see what the phone does with the out of focus areas to emulate bokeh.

Not interested. I just want to analyze resolving power for the scope of this thread. Bokeh rendering for phones is not something that I'm interested to test.

Then you have to use a camera mode that does not use all the sorcery tricks. Use the native focal length of the prime lens you are selecting. Likely that means 12.5MP binned mode and raw file in pro mode (or jpg in a raw container).

Because if you use any other focal length, the phone is likely to do it's sorcery and mix (fuse, blend, stretch, fit and crop) information from two lenses and two sensors and two resolutions together, which will likely also introduce some degree of fake detail artifacts. And your resolution results will only be representative for that very specific scenario.

And make sure the scene is bright enough, else the phone may decide to stack pictures regardless of what camera mode you are in (depends on make and model of the phone, in my S23U it still does computational processing and noise reduction in "expert raw" mode, but not in "pro" mode. Pro mode is most similar to how our m43 cameras work, but pics will be worse for "pro" mode in low light compared with "expert raw" mode where stacking is automatically used to reduce noise).

It's not that easy to compare resolution between an m43 camera with a fixed resolution, to a modern multi-camera phone with all this undocumented secret sorcery going on.

 

[HicHic]

Fabric but on a tilted plane. Then you see what the phone does with the out of focus areas to emulate bokeh.

Not interested. I just want to analyze resolving power for the scope of this thread. Bokeh rendering for phones is not something that I'm interested to test.

Then you have to use a camera mode that does not use all the sorcery tricks. Use the native focal length of the prime lens you are selecting. Likely that means 12.5MP binned mode and raw file in pro mode (or jpg in a raw container).

Because if you use any other focal length, the phone is likely to do it's sorcery and mix (fuse, blend, stretch, fit and crop) information from two lenses and two sensors and two resolutions together, which will likely also introduce some degree of fake detail artifacts. And your resolution results will only be representative for that very specific scenario.

And make sure the scene is bright enough, else the phone may decide to stack pictures regardless of what camera mode you are in (depends on make and model of the phone, in my S23U it still does computational processing and noise reduction in "expert raw" mode, but not in "pro" mode. Pro mode is most similar to how our m43 cameras work, but pics will be worse for "pro" mode in low light compared with "expert raw" mode where stacking is automatically used to reduce noise).

It's not that easy to compare resolution between an m43 camera with a fixed resolution, to a modern multi-camera phone with all this undocumented secret sorcery going on.

Yeah that's why I'm only using the 24mm for all these tests. Things will get messy fast with other zoom ranges. I actually have my Pixel phone set to disable the generic zoom ranges, and only allow me to choose from 1 of the 3 rear cameras to use. That way, I'm never accidentally digitally zooming in or mixing cameras together. I treat it like a real camera with a bag of 3 different primes, so it is a bit different from how most people use their phone cameras.