LX7 awful edge resolution due to distortion correction

Started Apr 30, 2013 | Discussions
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: Distortion is the best compromise
In reply to Samuel Dilworth, May 2, 2013

Samuel Dilworth wrote:

Of the three latest LX-series cameras, the LX7 has the worst wide-angle distortion by far – necessary to squeeze such a fast and sharp lens into such a small package (barely bigger than the LX5, but twice as bright). The LX5 has the least distortion, and the LX3 is in between, though closer to the LX5 than the LX7.

Going to 24 mm-e in itself is not the problem. The problem is making such a lens both very fast and very small in size (and cheap too!). The Sony RX100 only goes to 28 mm-e, but because it’s fast and extremely miniaturised it has very similar distortion to the LX7 at 24 mm-e (and it’s soft in the corners even before distortion correction).

Yes, these compacts are all a compromise, and I don't like the trade-off between lens size and image quality. I'm a huge fan of 24mm and when looking at images from recent compacts, those from the LX7 were probably still the best (most natural looking) in the corners that I could find - better than Canon S100/110 or even the 28mm setting from RX100.

It's not just the 'stretching' from distortion correction, but also the correction for vignetting etc. that produces all kinds of artifacts like increased noise and sometimes hue shift in the corners. LX7 does pretty well there. For really high corner quality in a compact the only option seems to be using prime lenses (e.g. Ricoh GR, unfortunately only 28mm effective). Seems there is not sufficient demand in the market for a compact camera with best possible image quality and a bit bigger lens ...

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: LX7 as a fish-eye lens
In reply to guyindasky, May 2, 2013

sure looks great for an image taken wide open at 24mm effective (but would need to see full size to be sure ...).

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: LX7 awful edge resolution due to distortion correction
In reply to Detail Man, May 2, 2013

Detail Man wrote:

I find DxO's rectiliear (barrel, pincushion) distortion corrections to be higher quality than the corrections that LR/CR, Silkpix, RAW Therapee, etc. perform using the Panasonic data. The corrections appear to based on a more complicated process which does more than correct from a single point in the center of the image-frame. I cannot replicate it using Lightroom or RAW Therapee. Have not tried, but strongly suspect the same is true where it comes to Silkypix.

I don't have experience with this combination, but in general PTLens also does a good job for distortion correction, resulting in much better corner sharpness than LR/CR or in-camera processing. Probably those more complicated distortion corrections cannot be done in-camera at the time

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
Kevin Coppalotti
Veteran MemberPosts: 8,126Gear list
Like?
discovering the contents of a sausage
In reply to Shirozina, May 2, 2013

If you look at the DXO display used by the OP it is pretty shocking, it i going to shake your confidence in the brilliance of leica optics, much akin to discovering what sausages are made of.

sometimes you can get too much information...lol.

-- hide signature --
Reply   Reply with quote   Complain
Cyril Catt
Veteran MemberPosts: 3,216
Like?
Re: LX7 awful edge resolution due to distortion correction
In reply to technic, May 2, 2013

technic wrote:

Cyril Catt wrote:

Less than a decade ago it was rare to find a p&s or bridge camera with a 24mm wide lens. Now there are many. But even now you cannot find a DSLR with an 18mm wide, F1.4 fast lens - certainly not at the same price.

True if you only look at nominal aperture, but if you look at equivalent aperture the situation is very different.

Samyang has a relatively good and cheap 2.8/14mm nominal for FF and an announced 2.8/10mm (15-16mm effective) for APS-C. Many of the APS-C SWA zooms like Tokina 2.8/11-16 or Canon 3.5/10-22 have a better equivalent aperture than the LX series (equivalent aperture for LX7 is about f/7 at the wide end), and they go as wide as the LX7 including the converter ...

You are right, technic. I would not expect the LX7 to outclass what could be achieved with a good DSLR. But Amazon currently has the 14mm Samyang at $379, the Tokina at $569, and the Canon at $646 to $949. So irrespective of what you mount them on, the combination will be bigger, heavier and more costly than the LX7 and converter. For those reasons, the LX7 no doubt attracts a different group of buyers, with different needs.

-- hide signature --

Cyril

Reply   Reply with quote   Complain
Sean Nelson
Forum ProPosts: 10,527
Like?
Re: Distortion is the best compromise
In reply to technic, May 2, 2013

technic wrote:

It's not just the 'stretching' from distortion correction...

You know that "stretching" in the corners is normal to get a rectilinear image at very wide angles, right?   It doesn't matter whether this is done in software or by the optical design - you can't have straight lines appearing straight and also have a very wide field of view without corner stretching.   Without that stretching you get lesser or greater degrees of barrel distortion which, if you go to the extreme, gives you a fisheye lens.

Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: LX7 awful edge resolution due to distortion correction
In reply to Cyril Catt, May 2, 2013

Maybe LX7 partly attracts the same type of photographers, despite the price and weight difference

I'm a DSLR user, but I'm very interested in a quality compact with good WA image quality. LX7 is a bit too big for my taste though, as a 'second camera' (not really pocketable, and no EVF which is a problem for me due to poor eyesight; so for me LF1 could be interesting despite the fact that it has no 24mm).

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: Distortion is the best compromise
In reply to Sean Nelson, May 2, 2013

Sean Nelson wrote:

technic wrote:

It's not just the 'stretching' from distortion correction...

You know that "stretching" in the corners is normal to get a rectilinear image at very wide angles, right?   It doesn't matter whether this is done in software or by the optical design - you can't have straight lines appearing straight and also have a very wide field of view without corner stretching.   Without that stretching you get lesser or greater degrees of barrel distortion which, if you go to the extreme, gives you a fisheye lens.

Wrong, it makes a HUGE difference if the stretching is done by optics or by software; at least as long as we have relatively low resolution sensors.

I'm not worried about distortion related to optical projection, or (not) having straight lines. I'm worried about the loss of sharpness in the corners, and especially the very asymmetric loss which often looks ugly/unnatural to me (pixels become elongated ellipses towards the corners).

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,270
Like?
Re: Distortion is the best compromise
In reply to technic, May 2, 2013

technic wrote:

I don't have experience with this combination, but in general PTLens also does a good job for distortion correction, resulting in much better corner sharpness than LR/CR or in-camera processing. Probably those more complicated distortion corrections cannot be done in-camera at the time.

Not familiar with what PTLens does either (the profile-data is collected by individual users from characteristics of their own individual units). If it uses the same process that the Panasonic image-file meta-data uses, I think that it is a fairly simple correction algorithm that proceeds from a single point located at the center of the image-frame, the identity for which is shown here:

http://en.wikipedia.org/wiki/Distortion_%28optics%29#Software_correction

From attempting to replicate what DxO Optics Pro does using RAW Therapee 4.0.9.x's rectlinear distortion correction tool, I am convinced that DxO corrections utilize more than a single reference-point located at the center of image-frame. I think DxO uses multiple reference-points.

Sean Nelson wrote:

technic wrote:

It's not just the 'stretching' from distortion correction...

You know that "stretching" in the corners is normal to get a rectilinear image at very wide angles, right?   It doesn't matter whether this is done in software or by the optical design - you can't have straight lines appearing straight and also have a very wide field of view without corner stretching.   Without that stretching you get lesser or greater degrees of barrel distortion which, if you go to the extreme, gives you a fisheye lens.

Wrong, it makes a HUGE difference if the stretching is done by optics or by software; at least as long as we have relatively low resolution sensors.

I'm not worried about distortion related to optical projection, or (not) having straight lines. I'm worried about the loss of sharpness in the corners, and especially the very asymmetric loss which often looks ugly/unnatural to me (pixels become elongated ellipses towards the corners).

I think that Sean is talking about what given amount of correction-processing must occur in software given whatever amount of optical rectilinear distortion exists as the result of particular lens design.

Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
PTLens
In reply to Detail Man, May 2, 2013

Detail Man wrote:

technic wrote:

I don't have experience with this combination, but in general PTLens also does a good job for distortion correction, resulting in much better corner sharpness than LR/CR or in-camera processing. Probably those more complicated distortion corrections cannot be done in-camera at the time.

Not familiar with what PTLens does either (the profile-data is collected by individual users from characteristics of their own individual units). If it uses the same process that the Panasonic image-file meta-data uses, I think that it is a fairly simple correction algorithm that proceeds from a single point located at the center of the image-frame, the identity for which is shown here:

PTLens does NOT use a simple algorithm but more complex functions which can fully correct moustache distortion (unlike the algorithms in LR/ACR and most other RAW converters and in-camera processing).

http://epaperpress.com/ptlens/distortion.html

However, AFAIK it uses fixed parameters for a certain lens (focal length) normally calibrated at infinity focus, and does not correct for changes in distortion with focus distance (potentially DXO can do that, but it's not clear to me if/which of their software really uses this).

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
LeRentier
Forum ProPosts: 13,622Gear list
Like?
Re: LX7 awful edge resolution due to distortion correction
In reply to Shirozina, May 2, 2013

Shirozina wrote:

The last image had some noise reduction applied to the processed image so here it is without and also the thumbnail showing what the distortion correction is actually doing.

Those shots are not even in focus.

You complain about the LX7 but your results are obtained via a few pieces of software which functioning you don't question.

Honestly, RTFM and learn how to focus before you declare a camera (any camera) is no good.

-- hide signature --

veni vidi vomi

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,270
Like?
Re: PTLens
In reply to technic, May 3, 2013

technic wrote:

Detail Man wrote:

technic wrote:

I don't have experience with this combination, but in general PTLens also does a good job for distortion correction, resulting in much better corner sharpness than LR/CR or in-camera processing. Probably those more complicated distortion corrections cannot be done in-camera at the time.

Not familiar with what PTLens does either (the profile-data is collected by individual users from characteristics of their own individual units). If it uses the same process that the Panasonic image-file meta-data uses, I think that it is a fairly simple correction algorithm that proceeds from a single point located at the center of the image-frame, the identity for which is shown here:

PTLens does NOT use a simple algorithm but more complex functions which can fully correct moustache distortion (unlike the algorithms in LR/ACR and most other RAW converters and in-camera processing).

http://epaperpress.com/ptlens/distortion.html

However, AFAIK it uses fixed parameters for a certain lens (focal length) normally calibrated at infinity focus, and does not correct for changes in distortion with focus distance (potentially DXO can do that, but it's not clear to me if/which of their software really uses this).

Not clear to me, either. Not so for the LX3, but for the LX5 and GH2 (becaue most Panasonic cameras do not include information estimating Focus Distance in their image-file meta-data), DxO has a control-slider for the user to input an estimated Focus Distance.

Have not seen the value of Focus Distance entered alter the Rectlinear Distortion correction. At certain specific "steps" between input Focus Distance values, the "Lens Softness" corrections cause the previewed image to shift the position of the previewed image slightly. Thus, I think that it may only be the "Lens Softness" corrections that are altered by Focus Distance estimate input by user.

Took a look at the demo here:

http://epaperpress.com/ptlens/lensCorrect.html

That looks interesting. The OP already uses LR, so that might be a more desirable option for them.

The only potential issue that I would see is that the profiles (similar to the Adobe LR/CR custom lens profiles) are derived from individual volunteer user test data submitted. Thus, they (may) not represent a good statistical sampling of the existing units (something that I assume that DxO Labs does do when creating their DxO Optical Corrections Modules). The other concern is that it is possible that those individual users (may) not have performed the testing properly.

That said, I am not trying promote DxO Optics Pro, I assume that you are not trying to promote PTLens (as it sounds like you have not yourself used it - at least in the case of the LX3, LX5, or LX7 ?), and "the proof of the pudding is in the eating" where it comes to actual results.

Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: PTLens
In reply to Detail Man, May 3, 2013

Detail Man wrote:

That said, I am not trying promote DxO Optics Pro, I assume that you are not trying to promote PTLens (as it sounds like you have not yourself used it - at least in the case of the LX3, LX5, or LX7 ?), and "the proof of the pudding is in the eating" where it comes to actual results.

I'm not promoting anything, just hoping that we will see higher quality in-camera corrections when it comes to this issue; to me the corner blurring from 'stretching' in current compact WA lenses is very annoying. PTLens is cheap but it will not fit everyone's workflow; it is years ago that I used it, mostly for distortion correction on converter lenses

Yes, the PTLens profiles will not always be perfect but you can adapt them if you want (lots of work though). I think DXO partly has the same problem: you can sample many lenses, but if there is significant variation in parameters only individual calibration will work if you want best results; we will probably get there in a few years.

I'm generally impressed with DXO lens corrections etc., but I don't think it suits my very basic workflow. But when I buy a new camera (currently looking for a quality WA compact and a new FF DSLR) I will look at it again.

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
Shirozina
Regular MemberPosts: 310
Like?
Re: LX7 awful edge resolution due to distortion correction
In reply to LeRentier, May 4, 2013

LeRentier wrote:

Shirozina wrote:

The last image had some noise reduction applied to the processed image so here it is without and also the thumbnail showing what the distortion correction is actually doing.

Those shots are not even in focus.

You complain about the LX7 but your results are obtained via a few pieces of software which functioning you don't question.

Honestly, RTFM and learn how to focus before you declare a camera (any camera) is no good.

-- hide signature --

veni vidi vomi

Yes they are in focus ( quite obviously). Sharpening is turned off in Capture One - suggest you learn the difference between the 2 before making such judgements;-)

Reply   Reply with quote   Complain
Shirozina
Regular MemberPosts: 310
Like?
Re: PTLens
In reply to technic, May 4, 2013

technic wrote:

Detail Man wrote:

That said, I am not trying promote DxO Optics Pro, I assume that you are not trying to promote PTLens (as it sounds like you have not yourself used it - at least in the case of the LX3, LX5, or LX7 ?), and "the proof of the pudding is in the eating" where it comes to actual results.

I'm not promoting anything, just hoping that we will see higher quality in-camera corrections when it comes to this issue; to me the corner blurring from 'stretching' in current compact WA lenses is very annoying. PTLens is cheap but it will not fit everyone's workflow; it is years ago that I used it, mostly for distortion correction on converter lenses

Yes, the PTLens profiles will not always be perfect but you can adapt them if you want (lots of work though). I think DXO partly has the same problem: you can sample many lenses, but if there is significant variation in parameters only individual calibration will work if you want best results; we will probably get there in a few years.

I'm generally impressed with DXO lens corrections etc., but I don't think it suits my very basic workflow. But when I buy a new camera (currently looking for a quality WA compact and a new FF DSLR) I will look at it again.

Any distortion correction software is going to create a similar amount of image 'smear' with this lens due to the amount of remapping needed to correct the distortion. Sample variation in manufacturing is likely to mean that any off the shelf correction profiles are approximate and for perfect geometrical correction you need to create a custom profile which at least you can do with LR.

Reply   Reply with quote   Complain
technic
Veteran MemberPosts: 8,632Gear list
Like?
Re: PTLens
In reply to Shirozina, May 4, 2013

Shirozina wrote:

Any distortion correction software is going to create a similar amount of image 'smear' with this lens due to the amount of remapping needed to correct the distortion.

WRONG! Just try it and you can see the very significant difference yourself. The standard corrections use a lower quality (probably faster) algorithm and have extra loss of resolution.

Sample variation in manufacturing is likely to mean that any off the shelf correction profiles are approximate and for perfect geometrical correction you need to create a custom profile which at least you can do with LR.

You can do that with PTLens; don't know about LR but I doubt it uses the more complex equations for distortion correction that PTLens uses. Can LR do a full correction of mustache distortion? The last version I tried was not able to do this, yet many compact, bright WA lenses have exactly this problem at the wide side. Most programs only correction the barrel part of the distortion (which is more pronounced), or some average between barrel and pincushion that is acceptable to most users but definitely not a full correction.

 technic's gear list:technic's gear list
Canon EOS 450D Canon EF 50mm f/1.8 II Canon EF 200mm f/2.8L II USM Canon EF 300mm f/4.0L IS USM Canon EF-S 15-85mm f/3.5-5.6 IS USM +5 more
Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,270
Like?
Re: PTLens
In reply to technic, May 4, 2013

technic wrote:

Shirozina wrote:

Any distortion correction software is going to create a similar amount of image 'smear' with this lens due to the amount of remapping needed to correct the distortion.

WRONG! Just try it and you can see the very significant difference yourself. The standard corrections use a lower quality (probably faster) algorithm and have extra loss of resolution.

I agree. A correction process that utilizes multiple reference-points (as opposed to only a single reference-point located in the center of the image-frame) would be able to do a better job of correcting edge/corner lens-system rectilinear distortion. I believe (from my experience in comparing it's results across the image-frame with the results of single reference-point based corrections) that is what DxO Optics Pro does in implementing corrections (and it appears that PTLens may also do).

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads