Distortion Correction; What Do You Think?

[K0d]

The latest Leica Panasonic M43 8mm 1.7 looks like a fisheye in RAW but is sharp and corrected in camera without appearing to suffer any degradation of image quality. Optically its the resulting image that counts and can now be the new standard as all the new EVF cameras do not need to have any compatibility with any optical viewfinder models or film cameras. This will eventually leave legacy film era lens mount systems like Leica M or Pentax K as the only systems needing rectilinear correction. As for being an optical purist would you post only RAW images just to avoid lens profiles or image enhancement?

 

[paul hofseth]

Ad the Q above about which optics are easily adaptable across camera systems, my own old Nikon Zeiss and Leitz stuff can easily be moved, as can the more recent Voigtländers . and some recent Chinese products . Especially useful if like me, you dislike autofocus and autoaperture mechanisms (fixing on random twigs and tending not to age with grace) I presume that Angenieux and Kinoptik do not qualify since they can no longer be bought new.

p.

 

[bclaff]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

I agree that focal length is defined on the optical axis and this doesn't change.

However, if you look at the image at a point off the axis, it more closely matches the image you would get from a different focal length lens (without distortion). That was what I meant. The usual meaning of focal length is defined at the centre of the image only.

Corrections technically change field of view, not focal length, but....

On most of Canon's PowerShot cameras, raw capture is only available thanks to CHDK . Thus, Canon never expected anyone would know about the "corrections" applied in their JPEGs. However, CHDK raws reveal that most of the PowerShots have zoom lenses that are roughly as marked on the long end, but are of much shorter focal length on the wide end. Why? Well, they heavily vignette and distort, so their field of view needs to be cropped to get an evenly-lit rectangle of the right proportions. The typical difference is roughly equivalent to calling a 24mm lens 28mm, and the camera silently upscales the cropped raw to a JPEG with the full marketing-claimed pixel count for the camera. I.e., Canon quotes the effective focal length, computed from the delivered angle of view in JPEGs. I very much doubt other manufacturers do much different.

CHDK exposes the fact that Canon's cameras commonly report bogus values for many metrics. For example, the user ISO settings are deliberately not mapped into the correct internal APEX96 values for computing exposure: the high ISO numbers are commonly off by nearly a full stop. In CHDK, this is known as "market" vs. "real" values for exposure parameters. My point isn't that Canon's being bad or somesuch, but that it is actually the industry norm to lie about these sorts of things either to look better for marketing purposes or to improve user intuition about things (e.g., half of 1/8s shutter speed is NOT 1/15s, but 1/16s, except at some point people apparently thought those oddly-rounded numbers more intuitive).

I don't normally watch for Canon PowerShot lens patents because I have no manufacturer diagrams for confirmation. But your post made me curious and so I think I tracked one (actually several) down.

Here's the Canon PowerShot SX710 HS as it appears in the PhotonsToPhotos interactive Optical Bench..

Here's the distortion (you expose this using the Distortion button)

W, M, and T stand for Wide, Mid, and Telephoto respectively.

About 16% barrel distortion at the wide end and essentially no distortion at medium and telephoto.

[If you play with this note that due to a math error I haven't tracked down some angles at the telephoto end are reported 720 degrees too high eg. 722.90 should be 2.90 , 710.87 should be -9.13 etc.; and there is a ray tracing error at the telephoto end that is probably related.]

--
Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

 

[ProfHankD]

I don't normally watch for Canon PowerShot lens patents because I have no manufacturer diagrams for confirmation. But your post made me curious and so I think I tracked one (actually several) down.

Here's the Canon PowerShot SX710 HS as it appears in the PhotonsToPhotos interactive Optical Bench..

Here's the distortion (you expose this using the Distortion button)

W, M, and T stand for Wide, Mid, and Telephoto respectively.

About 16% barrel distortion at the wide end and essentially no distortion at medium and telephoto.

Yup, and that's one of the higher-end PowerShots.

For example, here's the camera JPEG from an ELPH115IS:

And here's uncorrected processing of the actual raw that came from (available thanks to CHDK):

Notice that the vignetting is not centered. The centering of the PowerShot lens projection on the sensor is often quite poor and varies significantly across copies of the same model. Canon's correction seems to deal with that by being extra conservative rather than having the corrections individually calibrated. It's not too hard to get a significantly wider view by tuning the corrections to individual copies of the camera.

Of course, Canon never intended for anyone to be seeing the raw image data, and the JPEG images delivered by the camera are actually fairly good, so...

 

[Entropy512]

I realize that a group of producers can agree to maintain some corrections, but with optics that can be easily fitted to different camera bodies and no universal correction standards, a well corrected lens with a longer life is preferrable to a cheaper one dependent on correction if you intend to keep it for a long time.

p.

Please tell! Where are these "optics that can be easily fitted to different camera bodies"?

All my lenses are made to fit one particular range of cameras and no others.

Yeah. The only lenses I've seen that fit many bodies are T-mount - which have NO electronic controls whatsoever.

 

[Entropy512]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

 

[Tom Axford]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

 

[bclaff]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

 

[K0d]

Post removed for safety reasons

 

[Tom Axford]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

 

[Jack Hogan]

Hello Toothwalker, welcome to the forum. Do you happen to be this Toothwalker?

 

[bclaff]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

You can see an example here
Focal Length at Closest Focus - 105mm f2.8G IF-ED AF-S VR Micro-Nikkor

It's easiest to apply this extension tube method to removable lenses than can be manually focused (at least to infinity).

 

[Tom Axford]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

You can see an example here
Focal Length at Closest Focus - 105mm f2.8G IF-ED AF-S VR Micro-Nikkor

It's easiest to apply this extension tube method to removable lenses than can be manually focused (at least to infinity).

That method requires that you measure the image magnification, which requires you to measure the size of the image and the size of the object in order to compute the image magnification. That is exactly the same as the method I described, except that you are using close objects rather than distant ones.

 

[ProfHankD]

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

Accuracy of focal length reporting by a lens to the camera is actually a big deal for IBIS.

I created a little interactive tool to help people decide things like what one manual IBIS (sensor-movement in body image stabilization) setting to use for a manual zoom lens. Just set the sliders for min and max focal length to the same number to get the analysis of a non-zoom lens.

It turns out that you need to be pretty darn accurate in order to get much more than 3EV worth of correction. For a simple example, suppose the lens is really 100mm. Setting the IBIS to 105mm will actually limit IBIS to providing no more than 4.1EV worth of correction! Real focal lengths often vary from marketing values by a few percent, and that difference alone would prevent more than about 5EV IBIS correction.

Of course, IBIS correction doesn't work perfectly far off axis due to lens projection issues, but severe distortion will make it work a lot worse. Corrections applied in post can't fix errors in the IBIS movement model.

BTW, at least as of 2020 (I don't think they've changed, but I haven't checked), the CIPA methods for measurement of IBIS effectiveness were only measuring pitch and yaw correction on axis with a fixed shake waveform. They ignore roll, X, Y, and Z motions! Ok, everybody ignores Z, because movements in/out along the lens axis just change magnification in insignificant ways at normal shooting distances. However, roll is particularly common, especially with two-handed grips, and they ignore it partly because you don't see it on axis, but probably moreso because OIS can't correct it, and their testing was originally designed for OIS. CIPA also does not model how the camera is held nor camera-generated shake nor resonances, which make huge differences. In my opinion (as per my EI2020 paper, Characterization of camera shake ), the CIPA ratings don't correlate well with real-world IBIS performance, and any numbers much above 5EV correction are about as trustworthy as Volkswagon's old "clean diesel" performance numbers. Just assume most IBIS reliably gets no more than about 3EV of correction and you'll be fine. ;-)

 

[bclaff]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

You can see an example here
Focal Length at Closest Focus - 105mm f2.8G IF-ED AF-S VR Micro-Nikkor

It's easiest to apply this extension tube method to removable lenses than can be manually focused (at least to infinity).

That method requires that you measure the image magnification, which requires you to measure the size of the image and the size of the object in order to compute the image magnification. That is exactly the same as the method I described, except that you are using close objects rather than distant ones.

It's far easier to measure magnification accurately and to take multiple measurements with different tubes as I do. You also have the option of using the center portion to eliminate any false readings due to distortion.

So no, I don't think of them as equivalent.

 

[Tom Axford]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

You can see an example here
Focal Length at Closest Focus - 105mm f2.8G IF-ED AF-S VR Micro-Nikkor

It's easiest to apply this extension tube method to removable lenses than can be manually focused (at least to infinity).

That method requires that you measure the image magnification, which requires you to measure the size of the image and the size of the object in order to compute the image magnification. That is exactly the same as the method I described, except that you are using close objects rather than distant ones.

It's far easier to measure magnification accurately and to take multiple measurements with different tubes as I do. You also have the option of using the center portion to eliminate any false readings due to distortion.

So no, I don't think of them as equivalent.

It relies on exactly the same measurement, i.e. the size of the image. There is no fundamental difference between the image of a close object and an image of a distant object where the images are of the same size.

You always have the choice of using just the centre of the image to eliminate the effects of distortion, but then the image size must be small and the relative error becomes larger. The systematic errors introduced by lens distortion will work out the same whether you do it your way or mine.

 

[bclaff]

...

Distortion correction does not alter the focal length.

Yes, that's correct.

However, distortion does alter the focal length. T...

No. Toothwalk is right that focal length is not affected.

You are using focal length as a proxy for (observed) angle of view, the observed angle of view is affected. But focal length is an optical property that is not affected by and corrections.

And this is what Sony is changing - the number used as a proxy for FOV.

If you look at EXIF, you'll see that Sony has a "real" focal length number in the metadata in addition to the standard EXIF focal length tag, which is significantly shorter - Sony uses the "standard" tag as a proxy for view..

Indeed, if you are given an unknown lens and want measure its focal length, probably the simplest way to do it is to measure the image size of a very distant object that subtends a known angle at the lens. The focal length is simply the multiplier that relates image size to angle subtended by the object.

The formal definition of focal length requires you to measure the image size of an object on the optical axis that subjects an infinitesimally small angle. In practice, it is much more sensible to measure an object that fills a reasonable proportion of the image. But the measured focal length will then depend on the angles of the two ends of the object relative to the axis (unless the lens is distortion free).

There are more accurate ways, particularly for removable lenses.
Working backwards from the angle of view is not the typical procedure (and of course fails for fisheye lenses).

What method would you suggest for measuring the focal length?

You can see an example here
Focal Length at Closest Focus - 105mm f2.8G IF-ED AF-S VR Micro-Nikkor

It's easiest to apply this extension tube method to removable lenses than can be manually focused (at least to infinity).

That method requires that you measure the image magnification, which requires you to measure the size of the image and the size of the object in order to compute the image magnification. That is exactly the same as the method I described, except that you are using close objects rather than distant ones.

It's far easier to measure magnification accurately and to take multiple measurements with different tubes as I do. You also have the option of using the center portion to eliminate any false readings due to distortion.

So no, I don't think of them as equivalent.

It relies on exactly the same measurement, i.e. the size of the image. There is no fundamental difference between the image of a close object and an image of a distant object where the images are of the same size.

You always have the choice of using just the centre of the image to eliminate the effects of distortion, but then the image size must be small and the relative error becomes larger. The systematic errors introduced by lens distortion will work out the same whether you do it your way or mine.

I disagree. Unless you know the precise distance you aren't measuring size.
And measuring angle is (far) less precise than measuring magnification.
Furthermore my technique works at other than infinity focus and with otherthan rectilinear projections.

 

[Tom Axford]

I disagree.

Yes, we certainly disagree!

 

[Toothwalker]

Hello Toothwalker, welcome to the forum. Do you happen to be this Toothwalker?

Thank you. Yes, that is me.

 

[Jack Hogan]

Hello Toothwalker, welcome to the forum. Do you happen to be this Toothwalker?

Thank you. Yes, that is me.

Excellent, I visited your old site often when it was up, learning much in the process, always intuitive and clear. I miss it and look forward to when you'll have time to bring more of it back online. In the meantime, glad you dropped in.

Jack