G7X Distortion in RAW files vs RX100M3

[Jeff Fenske]

They designed a bright semi-fisheye lens (+/- 20-100mm zoom with very strong distortion) that thanks to software correction can be used as a normal 24-100 zoom, with some obvious compromise in image quality.

By "semi-fisheye," I'm guessing you're referring to what it does in wide angle, because we've seen that the uncorrected 100mm image is slightly pincushion.

It seems that maybe the RX100 II could have also done this if they tried to get 24mm too, instead of only 28-100mm.

I'd rather have a compromised 24mm than no 24mm at all. But it would be better for some of us had they not compromised 24mm so much, even if they had to increase the lens size or decrease the image circle (to then get the multi-aspect ratio advantages too).

When comparing to the RX100-1/2 one could say that this is indeed a faster version of that lens with extra 24-28mm reach, 1.5 stop faster, 100mm and better IS and without giving up anything important in optical quality. But the RX100-1/2 lens isn't very good to start with, the RX100-3 lens is a lot better optically thanks to the smaller zoom range. If you compare to that the compromises in the G7X become more important.

Yeah, "better IS!!" Extremely important, which is still often overlooked.

I'm definitely not interested in the Sony, but was just thinking that maybe they could have included the 24mm option into the II version, but didn't want to sacrifice IQ like Canon probably did.

I wonder why Canon didn't choose 25mm or 26mm wide, which could have helped IQ. Or better yet, they could offer smaller zoom steps, so we could choose 25 or 26mm, which I'm guessing we can't.

Perhaps they'll solve these issues and get even more zoom range with a curved sensor, someday.

Or they could produce a wide-angle-only zoom and tele-only zoom cameras to maximize IQ and even widen and lengthen the zoom range, which we discussed.

If some are willing to carry a fixed lens wide angle camera for maximum IQ, which manufacturers make, I would think this is a viable option.

Jeff Fenske, post: 54520622, member: 622952"]
Did Canon basically design a faster 28-100mm RX100 II lens, and extend its reach with software to get 24mm?

So 28-100mm will be fairly sharp, and 24mm is an added bonus, but will take a noticeable hit in image quality — if this does prove to be the case: 20MP from 15.26 MP!!

[/QUOTE]

 

[technic]

They designed a bright semi-fisheye lens (+/- 20-100mm zoom with very strong distortion) that thanks to software correction can be used as a normal 24-100 zoom, with some obvious compromise in image quality.

By "semi-fisheye," I'm guessing you're referring to what it does in wide angle, because we've seen that the uncorrected 100mm image is slightly pincushion.

yes, semi-fisheye in the lower part of the focal length range.

It seems that maybe the RX100 II could have also done this if they tried to get 24mm too, instead of only 28-100mm.

I would guess the RX100-1/2 lens is probably 25-26mm native, which is corrected into a 24mm equiv. nominal image. Most manufacturers still stick to old SLR prime focal length values for their zoom ranges, even when actual focal lengths are usually a bit different.

I'd rather have a compromised 24mm than no 24mm at all. But it would be better for some of us had they not compromised 24mm so much, even if they had to increase the lens size or decrease the image circle (to then get the multi-aspect ratio advantages too).

Corners of LX100 sample images don't look very good either at maximum WA, despite the small zoom range. That suggests that with the current technology and the lens dimensions that the camera companies are using, you either have to accept a smaller zoom range like in RX100-3, or a more compromised lens like in G7X. G7X - like RX100-1/2 - is more of a people/events camera where 'pixel-peeping details' are not important, while the RX100-3 is more suitable for landscape/architecture type photographer where you want maximum detail, good corners and as little artifacts as possible - within the possibilities of a very small camera.

You cannot extend the reach of a lens with software, if the data isn't there the software cannot invent it. The only thing the software can do is make the severely distorted native borders kind-off acceptable.

But could the data have been there in other cameras, but they considered the IQ to be too compromised, perhaps?

Could be, there are a few similar cameras that also throw away 25-30% of sensor pixels for the WA distortion correction. But difficult to tell what distortion patterns these cameras have without looking at the native RAW files. Would be an interesting subject for a comparative DPR Review ;-)

Again, I don't think G7X is doing anything fundamentally different that similar cameras, they are just balancing closer to the edge than in other lens designs.

When comparing to the RX100-1/2 one could say that this is indeed a faster version of that lens with extra 24-28mm reach, 1.5 stop faster, 100mm and better IS and without giving up anything important in optical quality. But the RX100-1/2 lens isn't very good to start with, the RX100-3 lens is a lot better optically thanks to the smaller zoom range. If you compare to that the compromises in the G7X become more important.

Yeah, "better IS!!" Extremely important, which is still often overlooked.

I'm definitely not interested in the Sony, but was just thinking that maybe they could have included the 24mm option into the II version, but didn't want to sacrifice IQ like Canon probably did.

The RX100-3 lens is a more advanced design than the 1/2 lens. I think Sony responded to the criticism about the 1/2 lens that was not quite up to the job. I don't know how well the RX100-3 sells, but judging from all the negative reactions on Sony forum regarding loss of 100mm it is clear than the average RX100 buyer gladly sacrifices image quality for more impressive specs.

I wonder why Canon didn't choose 25mm or 26mm wide, which could have helped IQ. Or better yet, they could offer smaller zoom steps, so we could choose 25 or 26mm, which I'm guessing we can't.

I guess they could have offered 21mm wide if they wanted, with even worse image quality ;-(

Perhaps they'll solve these issues and get even more zoom range with a curved sensor, someday.

Curved sensors are great for wide/bright primes, but don't help very much for wide-to-tele zoom lenses. The best solution is a bigger lens, but maybe it needs to be 2x bigger (longer) for a significant step up in quality, which is probably more than the average buyer will accept. Further into the future this problem could be solved with negative refractive index materials.

Or they could produce a wide-angle-only zoom and tele-only zoom cameras to maximize IQ and even widen and lengthen the zoom range, which we discussed.

If some are willing to carry a fixed lens wide angle camera for maximum IQ, which manufacturers make, I would think this is a viable option.

I like the idea because it would be much easier to make an optically very high quality lens this way. But I don't think the market accepts it, most consumers 'want it all and want it now!'.

 

[Ketan G]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Here is the full-resolution shot at 24mm, f/4:


24mm, f/4



I then isolated a number of 100% crops. While corners are always a bit softer, I was thrown by the difference between the center left (sharp) and center right (much softer). It seems that anything outside of the central region gets quite soft on my copy:


24mm f/4



As a point of comparison, the lens is incredibly sharp midway through the range, and even at full telephoto (100mm, f/2.8). They are sharpest shots I have seen from a compact and the sharpest I've seen with the Sony 1" sensor, beating the FZ1000, RX100, and RX10 which I own or have owned. Furthermore, it handles like a dream compared to the RX100 and RX10, and is as good or better than the FZ1000:


~50mm, f/4




100mm, f/2.8



--

Ketan Gajria Photography

Ketan Gajria Photography featuring pictures from more than 35 countries including Syria, Bhutan, Pakistan, Iraq, Kyrgyzstan, Jordan, Lebanon, Russia, China, Turkey, Indonesia, India, Japan, Nicaragua, Oman, and more.

www.ketangajria.com

 

[2eyesee]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Here is the full-resolution shot at 24mm, f/4:


24mm, f/4

You have to be careful with brick wall shots with cameras like these. The extreme fish-eye distortion (in uncorrected images) results in a curved focal plane, so you can't be sure everything is in focus even on a flat wall. Furthermore, you're pointing the camera up and it doesn't even look like you're square on.

If you can find a target that you can stand further back and square on, you may like to compare your results with my RX100M3 brick wall test. The RX100M3 is pretty soft at 24mm too, but sharpens up significantly at 28mm. I'd be interested to hear how the G7X improves at 28mm, as I'd probably end up using it more as a 28-100mm , and I like the way the lens stays fast at 28mm.

As a point of comparison, the lens is incredibly sharp midway through the range, and even at full telephoto (100mm, f/2.8). They are sharpest shots I have seen from a compact and the sharpest I've seen with the Sony 1" sensor, beating the FZ1000, RX100, and RX10 which I own or have owned. Furthermore, it handles like a dream compared to the RX100 and RX10, and is as good or better than the FZ1000:


~50mm, f/4


100mm, f/2.8

This 100mm f/2.8 looks nice and sharp - I would be happy with that! To me the headline feature of the G7X is the 100mm f/2.8, but I've seen a lot of very soft shots using this. I'm very happy to hear you're getting sharp images wide open at full tele though.

 

[technic]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Agree about soft corners but this isn't unusual for similar cameras at 24mm equiv. (e.g. RX100-1/2, Canon S100/110, Powershot G1X II, even LX100). Only RX100-3 looks clearly better, maybe there are others but in general it seems all these extra-compact lenses have soft corners especially in the WA range.

I'm more worried about the severe artifacts that occur in the corners with OOC JPEG shots (and which I also see in most images converted from RAW - not in your shot but it would be difficult to see with this subject). Looks to me like the PP that is required to make the corners look good from a distance is tied to increased noise (not a bit but really, there is several stops of extra noise in the corners) and other strange artifacts.

I then isolated a number of 100% crops. While corners are always a bit softer, I was thrown by the difference between the center left (sharp) and center right (much softer). It seems that anything outside of the central region gets quite soft on my copy:

agree this doesn't look good, there is just a small region that is really sharp; at f/4 the result should be better. Like 2eyesee mentions below, this could be a combination of field curvature,camera tilt and exact focusing point. I would recommend to try this with a subject at bigger distance (at least 10 meters or so) and where the optical axis is really at a right angle to the wall (and not tilted).

As a point of comparison, the lens is incredibly sharp midway through the range, and even at full telephoto (100mm, f/2.8). They are sharpest shots I have seen from a compact and the sharpest I've seen with the Sony 1" sensor, beating the FZ1000, RX100, and RX10 which I own or have owned.

Just theoretically (because of the required distortion corrections) one would expect the G7X to have the best center sharpness, and the worst corner sharpness ... I think quality of 100mm f/2.8 images strongly depends on subject distance (worse for close up) and subject contrast (much worse with high contrast details). RX100-3 clearly goes for relatively even sharpness across the frame, at the cost of slightly lower center sharpness (and smaller zoom range).

Also, there might be some sample variation between individual lenses e.g. at what focal length they perform best.

 

[Jeff Fenske]

They designed a bright semi-fisheye lens (+/- 20-100mm zoom with very strong distortion) that thanks to software correction can be used as a normal 24-100 zoom, with some obvious compromise in image quality.

By "semi-fisheye," I'm guessing you're referring to what it does in wide angle, because we've seen that the uncorrected 100mm image is slightly pincushion.

yes, semi-fisheye in the lower part of the focal length range.

It seems that maybe the RX100 II could have also done this if they tried to get 24mm too, instead of only 28-100mm.

I would guess the RX100-1/2 lens is probably 25-26mm native, which is corrected into a 24mm equiv. nominal image. Most manufacturers still stick to old SLR prime focal length values for their zoom ranges, even when actual focal lengths are usually a bit different.

I'd rather have a compromised 24mm than no 24mm at all. But it would be better for some of us had they not compromised 24mm so much, even if they had to increase the lens size or decrease the image circle (to then get the multi-aspect ratio advantages too).

Corners of LX100 sample images don't look very good either at maximum WA, despite the small zoom range. That suggests that with the current technology and the lens dimensions that the camera companies are using, you either have to accept a smaller zoom range like in RX100-3, or a more compromised lens like in G7X. G7X - like RX100-1/2 - is more of a people/events camera where 'pixel-peeping details' are not important, while the RX100-3 is more suitable for landscape/architecture type photographer where you want maximum detail, good corners and as little artifacts as possible - within the possibilities of a very small camera.

The LX100 has bigger pixels, so their lens may smaller than ideal too, having compromised image quality, even with the decreased zoom range.

I'd like to see what Panasonic could do with a 1-inch sensor for the LX8, if they make one, hopefully.

You cannot extend the reach of a lens with software, if the data isn't there the software cannot invent it. The only thing the software can do is make the severely distorted native borders kind-off acceptable.

But could the data have been there in other cameras, but they considered the IQ to be too compromised, perhaps?

Could be, there are a few similar cameras that also throw away 25-30% of sensor pixels for the WA distortion correction. But difficult to tell what distortion patterns these cameras have without looking at the native RAW files. Would be an interesting subject for a comparative DPR Review ;-)

Again, I don't think G7X is doing anything fundamentally different that similar cameras, they are just balancing closer to the edge than in other lens designs.

When comparing to the RX100-1/2 one could say that this is indeed a faster version of that lens with extra 24-28mm reach, 1.5 stop faster, 100mm and better IS and without giving up anything important in optical quality. But the RX100-1/2 lens isn't very good to start with, the RX100-3 lens is a lot better optically thanks to the smaller zoom range. If you compare to that the compromises in the G7X become more important.

Yeah, "better IS!!" Extremely important, which is still often overlooked.

I'm definitely not interested in the Sony, but was just thinking that maybe they could have included the 24mm option into the II version, but didn't want to sacrifice IQ like Canon probably did.

The RX100-3 lens is a more advanced design than the 1/2 lens. I think Sony responded to the criticism about the 1/2 lens that was not quite up to the job. I don't know how well the RX100-3 sells, but judging from all the negative reactions on Sony forum regarding loss of 100mm it is clear than the average RX100 buyer gladly sacrifices image quality for more impressive specs.

The loss of 100mm is huge. 70mm is hardly even a telephoto.

I shot extensively with the LX5, which is 24-90mm. 90mm isn't much different than 100mm, and is far better than 70 or 75 (LX100). I could live with 90mm.

I also shot with a Canon G10 a lot, which was 28-140mm at 15MP — awesome camera for the day, except in low light, and it didn't have 24mm wide! 140mm was a blast, and I used it a lot! It would be much harder to do with a 1-inch sensor, though.

I wonder why Canon didn't choose 25mm or 26mm wide, which could have helped IQ. Or better yet, they could offer smaller zoom steps, so we could choose 25 or 26mm, which I'm guessing we can't.

I guess they could have offered 21mm wide if they wanted, with even worse image quality ;-(

Perhaps they'll solve these issues and get even more zoom range with a curved sensor, someday.

Curved sensors are great for wide/bright primes, but don't help very much for wide-to-tele zoom lenses. The best solution is a bigger lens, but maybe it needs to be 2x bigger (longer) for a significant step up in quality, which is probably more than the average buyer will accept. Further into the future this problem could be solved with negative refractive index materials.

I have heard that curved don't work for long teles, but thought they might with 100mm or so.

I hauled around on my belt the Canon G10, 24/7, and I'd be willing to carry LX100 sized camera if the features and image quality were there.

It's exciting to have great photographic potential always at the ready.

Or they could produce a wide-angle-only zoom and tele-only zoom cameras to maximize IQ and even widen and lengthen the zoom range, which we discussed.

If some are willing to carry a fixed lens wide angle camera for maximum IQ, which manufacturers make, I would think this is a viable option.

I like the idea because it would be much easier to make an optically very high quality lens this way. But I don't think the market accepts it, most consumers 'want it all and want it now!'.

 

[John TF]

Again, I don't think G7X is doing anything fundamentally different that similar cameras, they are just balancing closer to the edge than in other lens designs.

+1 That's my impression, too. They seem to have ventured out further into the margins, for marketing reasons obviously, but also perhaps for those marginal shooting situations where it is better to have imperfect rendition than no image at all. Therefore I would treat the realistic focal range as 24/2.8 to 100/3.5, with 24/1.8 and 100/2.8 reserved as "extra add-on" values to be used only in really low light when visual acuity is not particularly sharp in any case, so their imperfections might be more acceptable. In the same way, that extra add-on ISO value (12,800 or whatever) that all the cameras seem to have isn't really intended to be taken seriously - it is just an extreme fall-back position. Same with this lens, IMHO.

 

[Ketan G]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Here is the full-resolution shot at 24mm, f/4:


24mm, f/4

You have to be careful with brick wall shots with cameras like these. The extreme fish-eye distortion (in uncorrected images) results in a curved focal plane, so you can't be sure everything is in focus even on a flat wall. Furthermore, you're pointing the camera up and it doesn't even look like you're square on.

If you can find a target that you can stand further back and square on, you may like to compare your results with my RX100M3 brick wall test. The RX100M3 is pretty soft at 24mm too, but sharpens up significantly at 28mm. I'd be interested to hear how the G7X improves at 28mm, as I'd probably end up using it more as a 28-100mm , and I like the way the lens stays fast at 28mm.

I agree with your general points, but am fairly confident this is accurate for the following reasons:

-8.8mm f/4 should be equivalent to 24mm f/11 on full-frame. Having shot that on full-frame, I know that pretty much everything in the frame, even accounting for the increased subject distance of the corners, is in focus. Were there extreme pincushion I could see something "closer" to the lens being out of focus, but the corners are actually further back, not closer.

-I also shot at f/1.8, 2.8, and 5.6, the last of which is equivalent to f/16. There was a slight improvement at apertures narrower than f/1.8 but no change after that.

-Although we can take into account diffraction, my main point was not about the corners, but even the center-right, which is definitely in focus from wide open on down.

As a point of comparison, the lens is incredibly sharp midway through the range, and even at full telephoto (100mm, f/2.8). They are sharpest shots I have seen from a compact and the sharpest I've seen with the Sony 1" sensor, beating the FZ1000, RX100, and RX10 which I own or have owned. Furthermore, it handles like a dream compared to the RX100 and RX10, and is as good or better than the FZ1000:


~50mm, f/4


100mm, f/2.8

This 100mm f/2.8 looks nice and sharp - I would be happy with that! To me the headline feature of the G7X is the 100mm f/2.8, but I've seen a lot of very soft shots using this. I'm very happy to hear you're getting sharp images wide open at full tele though.

--

Ketan Gajria Photography

Ketan Gajria Photography featuring pictures from more than 35 countries including Syria, Bhutan, Pakistan, Iraq, Kyrgyzstan, Jordan, Lebanon, Russia, China, Turkey, Indonesia, India, Japan, Nicaragua, Oman, and more.

www.ketangajria.com

 

[photofisher]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Here is the full-resolution shot at 24mm, f/4:


24mm, f/4

I then isolated a number of 100% crops. While corners are always a bit softer, I was thrown by the difference between the center left (sharp) and center right (much softer). It seems that anything outside of the central region gets quite soft on my copy:


24mm f/4

As a point of comparison, the lens is incredibly sharp midway through the range, and even at full telephoto (100mm, f/2.8). They are sharpest shots I have seen from a compact and the sharpest I've seen with the Sony 1" sensor, beating the FZ1000, RX100, and RX10 which I own or have owned. Furthermore, it handles like a dream compared to the RX100 and RX10, and is as good or better than the FZ1000:


~50mm, f/4


100mm, f/2.8

--
http://www.ketangajria.com

Thanks for posting. How about at 28mm?

 

[Jeff Fenske]

Perhaps they'll solve these issues and get even more zoom range with a curved sensor, someday.

Curved sensors are great for wide/bright primes, but don't help very much for wide-to-tele zoom lenses. The best solution is a bigger lens, but maybe it needs to be 2x bigger (longer) for a significant step up in quality, which is probably more than the average buyer will accept. Further into the future this problem could be solved with negative refractive index materials.

I think you're right saying: "The best solution is a bigger lens."

The 1-inch sensor is much bigger than the 1/1.7-inch sensors, so I would think the lenses need to be that much bigger too in order for the lens not to cripple the image quality of the camera, overall.

My main concern is the purple fringing, which we think is CA, that is also showing up at 50mm. From what I've seen in 35mm FF lens tests, this indicates a lack of contrast in the lens overall. So the image quality is muted, the light doesn't get through crisp and clean. Vividness and clarity suffer, and probably the resolution too.

I'm mostly familiar with Canon FF lenses, but this is probably true for Nikon and others as well. The FF zooms have a lot of character and clarity now — contrast. Few photographers feel the need to shoot primes anymore in FF. But this isn't yet true in M43, where the primes seem to have more contrast. [Perhaps some of the latest lenses are changing this. I haven't checked recently.]

It seems that when Canon (and probably the others) upgraded their zoom lenses in the last 10 years or more, they seemed to almost always increase the filter size without changing aperture — from 77mm to 82mm, for example. So they increased the diameter of the elements without changing the fastest f/stop or focal length. And the specs and vividness are fantastic!

Another example is the new 50 and 55mm lenses from Sigma and Zeiss, with their state of the art specs. They both have 77mm filter sizes, while the Canon and Nikon 50mm f/1.4s have 58mm filter sizes.

In M43, Olympus and Panasonic have been so interested in keeping the lenses small that they've probably been crippling their zoom lenses' IQ, trying to keep them so small.

So perhaps in M43 and in some of these compacts, the lenses need to be made larger in diameter in order to bring in more light and minimize CA, to get greatly improved contrast, so the lenses no longer cripple the image quality of the 1-inch sensor.

And by decreasing the image circle slightly, incorporating Panny's multi-aspect feature onto a 1-inch sensor, the lens won't have to be as large as if they keep the image circle the same.

 

[Ketan G]

Unfortunately I don't have that, I'm currently trying to return the camera.

 

[2eyesee]

Unfortunately I'm seeing the same issues. I shot a wall at 24mm using a sturdy tripod, self timer, RAW, zero NR in DPP, then sharpened extensively in Photoshop but with a low radius to show where the lens is sharp. I shot at all apertures and was extremely disappointed.

Here is the full-resolution shot at 24mm, f/4:


24mm, f/4

You have to be careful with brick wall shots with cameras like these. The extreme fish-eye distortion (in uncorrected images) results in a curved focal plane, so you can't be sure everything is in focus even on a flat wall. Furthermore, you're pointing the camera up and it doesn't even look like you're square on.

If you can find a target that you can stand further back and square on, you may like to compare your results with my RX100M3 brick wall test. The RX100M3 is pretty soft at 24mm too, but sharpens up significantly at 28mm. I'd be interested to hear how the G7X improves at 28mm, as I'd probably end up using it more as a 28-100mm , and I like the way the lens stays fast at 28mm.

I agree with your general points, but am fairly confident this is accurate for the following reasons:

-8.8mm f/4 should be equivalent to 24mm f/11 on full-frame. Having shot that on full-frame, I know that pretty much everything in the frame, even accounting for the increased subject distance of the corners, is in focus. Were there extreme pincushion I could see something "closer" to the lens being out of focus, but the corners are actually further back, not closer.

-I also shot at f/1.8, 2.8, and 5.6, the last of which is equivalent to f/16. There was a slight improvement at apertures narrower than f/1.8 but no change after that.

-Although we can take into account diffraction, my main point was not about the corners, but even the center-right, which is definitely in focus from wide open on down.

I would still be careful about drawing too many conclusions based on shots like these. The G7X is no ordinary 24mm - it goes into uncharted territory for the extreme barrel distortion it is correcting for.

I understand you are in the process of returning the camera, but did you notice it improve significantly at 28mm? I would expect the distortion correction to me much less, so optical performance should improve.