What is mFT FoV?

[gollywop]

If one has, say, a 12mm lens on an mFT camera, it presumably implies some FoV. However, these days a large number of mFT lenses do software correction for lens distortions, and the image as rendered in, say, ACR is significantly different from the uncorrected version of the same image.

My question: is the 12mm FoV relevant to the FoV of the original image or the FoV of the corrected image, which is, typically, significantly smaller than that of the original?

--
gollywop

 

[Detail Man]

gollywop wrote:

Detail Man wrote:

However, since correcting for barrel-distortion has the apparent effect of "squeezing" image-data outwards (beyond the horizontal borders of some given aspect-ratio), could it be that the inverse operation of correcting for pincushion-distortion (pincusion-distortion correction) would have an effect of "squeezing" image-data inwards (within the horizontal borders of some given aspect-ratio) - thus having the effect of actually decreasing the horizontal (and thus the diagonal) FOV existing in (pincusion-distortion corrected) image ?

Well, the pin-cushion correction could also be by pulling data outwards, but along the middles of the borders rather than the corners. If that's what happens (and it's my suspicion that it is), the "pulled-out" portion can simply be discarded and the original aspect and size retained. But I have no examples to analyze.

But have a look at the example mathematical identities set forth in this information source:

http://en.wikipedia.org/wiki/Distortion_(optics)#Software_correction


Radial distortion ... [is] ... primarily dominated by low order radial components ...

where:

(xd, yd) = distorted image point as projected on image plane using specified lens;

(xu, yu) = undistorted image point as projected by an ideal pin-hole camera;

(xc, yc) = distortion center (assumed to be the principal point);

Kn = Nth radial distortion coefficient;

and r equals the following:

Barrel distortion typically will have a positive term for K1 whereas pincushion distortion will have a negative value.

I take it from the above statement that barrel-distortion correction coefficients will have negative terms for Kn, and that pincusion-distortion correction coefficients will have positive terms for Kn.

Thus, it seems that the x-axis locations of corrected pixel-locations are shifted in opposite directions in the corrected images - causing pincushion-distortion corrections to push the x-axis location of pixels inwards, and rectilinear-distortion corrections to push the x-axis location of pixels outwards. Upon having a look at the above-quoted information, does that make sense to you ?

.

This matter seems potentially quite relevant to your original question posed in this thread.

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

However, it seem (to me) that a different situation appears to exist in the case of the existence of pincushion-distortion. In those cases, what I am thinking is that additional FOV can only be relalized (in the case of the existence of finite amounts of pincushion-distortion) when one is willing to decrease the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to decrease relative to the vertical dimension).

Given that today's preferential trends for image-presentation seem to in many cases be moving towards higher valued (as opposed to lower valued) aspect-ratios ("wider" displays and screens), it seems that RAW-images recorded using a telephoto zoom lens that (over some range of Focal Lengths of adjustment) exhibits pincusion-distortion appear to operate at a relative disadvantage (where realizable FOV is concerned) relative to the situation where (over some range of Focal Lengths of adjustment) barrel-distortion is exhibited.

(Perhaps) this is one reason why (at least to my limited knowledge regarding such leneses) fixed Focal Length lenses do not exhibit pincushion-distortion in their un-corrected image output ?

DM ...

 

[Lights]

gollywop wrote:

If one has, say, a 12mm lens on an mFT camera, it presumably implies some FoV. However, these days a large number of mFT lenses do software correction for lens distortions, and the image as rendered in, say, ACR is significantly different from the uncorrected version of the same image.

My question: is the 12mm FoV relevant to the FoV of the original image or the FoV of the corrected image, which is, typically, significantly smaller than that of the original?

--
gollywop

I don't know if it can be answered, since so many conversion programs seem to convert so differently in regard to size. I know my Panasonic 14 with my WA conversion lens and a spacer or two, will vignette with some Raw converters and not with others (and not with jpg). A good question however, since it makes a person realize, and be aware that it seems there isn't a standard, or at least a standard that we're privy to.

--
My Gallery is here -
http://www.pbase.com/madlights

Why so serious? :The Joker

 

[gollywop]

Detail Man wrote:

gollywop wrote:

Detail Man wrote:

However, since correcting for barrel-distortion has the apparent effect of "squeezing" image-data outwards (beyond the horizontal borders of some given aspect-ratio), could it be that the inverse operation of correcting for pincushion-distortion (pincusion-distortion correction) would have an effect of "squeezing" image-data inwards (within the horizontal borders of some given aspect-ratio) - thus having the effect of actually decreasing the horizontal (and thus the diagonal) FOV existing in (pincusion-distortion corrected) image ?

Well, the pin-cushion correction could also be by pulling data outwards, but along the middles of the borders rather than the corners. If that's what happens (and it's my suspicion that it is), the "pulled-out" portion can simply be discarded and the original aspect and size retained. But I have no examples to analyze.

But have a look at the example mathematical identities set forth in this information source:

http://en.wikipedia.org/wiki/Distortion_(optics)#Software_correction

Radial distortion ... [is] ... primarily dominated by low order radial components ...

where:

(xd, yd) = distorted image point as projected on image plane using specified lens;

(xu, yu) = undistorted image point as projected by an ideal pin-hole camera;

(xc, yc) = distortion center (assumed to be the principal point);

Kn = Nth radial distortion coefficient;

and r equals the following:

Barrel distortion typically will have a positive term for K1 whereas pincushion distortion will have a negative value.

I take it from the above statement that barrel-distortion correction coefficients will have negative terms for Kn, and that pincusion-distortion correction coefficients will have positive terms for Kn.

Thus, it seems that the x-axis locations of corrected pixel-locations are shifted in opposite directions in the corrected images - causing pincushion-distortion corrections to push the x-axis location of pixels inwards, and rectilinear-distortion corrections to push the x-axis location of pixels outwards. Upon having a look at the above-quoted information, does that make sense to you ?

.

This matter seems potentially quite relevant to your original question posed in this thread.

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

However, it seem (to me) that a different situation appears to exist in the case of the existence of pincushion-distortion. In those cases, what I am thinking is that additional FOV can only be relalized (in the case of the existence of finite amounts of pincushion-distortion) when one is willing to decrease the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to decrease relative to the vertical dimension).

Given that today's preferential trends for image-presentation seem to in many cases be moving towards higher valued (as opposed to lower valued) aspect-ratios ("wider" displays and screens), it seems that RAW-images recorded using a telephoto zoom lens that (over some range of Focal Lengths of adjustment) exhibits pincusion-distortion appear to operate at a relative disadvantage (where realizable FOV is concerned) relative to the situation where (over some range of Focal Lengths of adjustment) barrel-distortion is exhibited.

(Perhaps) this is one reason why (at least to my limited knowledge regarding such leneses) fixed Focal Length lenses do not exhibit pincushion-distortion in their un-corrected image output ?

DM ...

A touch of casual empiricism:

If I open an image of a rectangular item (in this case a CC24 card) whose edges closely border the edge of the frame in ACR and apply negative rectilinear distortion (pincushion) correction, it bows the edges out at their center at a much greater rate than it moves the corners. In fact, the corners barely move, even when the edges bow out considerably. That's what I would expect. The corner-to-corner (both vertically and horizontally) and the diagonal pixel count remains unchanged, but a horizontal (vertical) line in the center is stretched considerably outside the frame.

And, if one now crops the resulting "corrected" figure, the pixel dimensions are exactly the same as the initial (uncorrected) image.

This contrasts to the case of barrel distortion, where the corners again remain fixed, but the edges are bowed inwards. An "interior" crop now is necessarily smaller than the original in both dimensions, and, pari passu, the diagonal.

--
gollywop

 

[Detail Man]

Detail Man wrote:

gollywop wrote:

Detail Man wrote:

However, since correcting for barrel-distortion has the apparent effect of "squeezing" image-data outwards (beyond the horizontal borders of some given aspect-ratio), could it be that the inverse operation of correcting for pincushion-distortion (pincusion-distortion correction) would have an effect of "squeezing" image-data inwards (within the horizontal borders of some given aspect-ratio) - thus having the effect of actually decreasing the horizontal (and thus the diagonal) FOV existing in (pincusion-distortion corrected) image ?

Well, the pin-cushion correction could also be by pulling data outwards, but along the middles of the borders rather than the corners. If that's what happens (and it's my suspicion that it is), the "pulled-out" portion can simply be discarded and the original aspect and size retained. But I have no examples to analyze.

But have a look at the example mathematical identities set forth in this information source:

http://en.wikipedia.org/wiki/Distortion_(optics)#Software_correction


Radial distortion ... [is] ... primarily dominated by low order radial components ...

where:

(xd, yd) = distorted image point as projected on image plane using specified lens;

(xu, yu) = undistorted image point as projected by an ideal pin-hole camera;

(xc, yc) = distortion center (assumed to be the principal point);

Kn = Nth radial distortion coefficient;

and r equals the following:

Barrel distortion typically will have a positive term for K1 whereas pincushion distortion will have a negative value.

I take it from the above statement that barrel-distortion correction coefficients will have negative terms for Kn, and that pincusion-distortion correction coefficients will have positive terms for Kn.

Thus, it seems that the x-axis locations of corrected pixel-locations are shifted in opposite directions in the corrected images - causing pincushion-distortion corrections to push the x-axis location of pixels inwards ...

... outward ...

... and rectilinear-distortion corrections to push the x-axis location of pixels outwards ...

... inward.

Upon having a look at the above-quoted information, does that make sense to you ?

It also appears that the y-axis locations of corrected pixel-locations are shifted in opposite directions in the corrected images - causing pincushion-distortion corrections to push the y-axis location of pixels outward, and rectilinear-distortion corrections to push the y-axis location of pixels inward.

Both x and y co-ordinates of the image projected by the lens-system onto the image-sensor are "shrunk" in the case of the correction of barrel-distortion (allowing for the effective FOV to be increased when the image-data is represented within some given image pixel-size).

The x and y co-ordinates of the image projected by the lens-system onto the image-sensor are "expanded" in the case of the correction of pincushion-distortion (having the effect of causing the effective FOV to be decreased when the image-data is represented within some given image pixel-size).

This matter seems potentially quite relevant to your original question posed in this thread.

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

However, it seem (to me) that a different situation appears to exist in the case of the existence of pincushion-distortion.

The difference appears to be that the existence of corrections being necessary for finite amounts of pincushion-distortion represent a reduction in the effective FOV that can as a result be realized when the image-data is represented within some given output image pixel-size.

In those cases, what I am thinking is that additional FOV can only be relalized (in the case of the existence of finite amounts of pincushion-distortion) when one is willing to decrease the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to decrease relative to the vertical dimension).

Given that today's preferential trends for image-presentation seem to in many cases be moving towards higher valued (as opposed to lower valued) aspect-ratios ("wider" displays and screens),

It seems that RAW-images recorded using a telephoto zoom lens that (over some range of Focal Lengths of adjustment) exhibits pincusion-distortion appear to operate at a relative disadvantage (where realizable FOV is concerned) relative to the situation where (over some range of Focal Lengths of adjustment) barrel-distortion is exhibited.

(Perhaps) this is one reason why (at least to my limited knowledge regarding such leneses) fixed Focal Length lenses do not exhibit pincushion-distortion in their un-corrected image output ?

DM ...

 

[Guy Parsons]

I'm munching lunch right now, but when finished I promise that I'll find my pincushionest lens and do the usual animated gif of uncorrected jpeg and out of camera jpeg - that way we'll see what Olympus does to the corners and edges etc.

Back soon..... regards..... Guy

 

[Guy Parsons]

Here's the Mk1 Oly 14-42 at 42mm on E-PL5, seems to be about the most pincushion I can muster in the stable.

Out of camera jpeg 4608 x 3456, out of Picasa uncorrected jpeg 4640 x 3472.

No attempt to make the two look the same, the out of camera jpeg is wishy-washy for other reasons and I haven't bothered to fine tune it just for this example.

The Picasa jpeg has the better sky etc. Never mind the quality, feel the width. Both resized to 560 x 420 for the gif.




Not sure what it proves, but hey, it's fun to see it unbend like this.

Regards...... Guy

 

[Guy Parsons]

Whilst fiddling with camera defaut jpeg and then from Oly Viewer 2 and from Silkypix Pro V5 - all delivering 4608 x 3456 jpegs we see these differences......(resized to 560 x 420)

First is camera jpeg and OV2 jpeg with the camera one deliberately left pale and other made a bit darker .....



So Oly Viewer 2 does NOT do exactly what the camera does to make a jpeg.

Next is Oly Viewer 2 and Silkypix, again the Silkypix deliberately darker than the OV2 jpeg......



So even though all three make a corrected 4608 x 3456 jpeg, they are all a slightly different frame. Back to my original thoughts here, "how long is a piece of string?"


Regards...... Guy

 

[Detail Man]

Guy Parsons wrote:

Here's the Mk1 Oly 14-42 at 42mm on E-PL5, seems to be about the most pincushion I can muster in the stable.

Out of camera jpeg 4608 x 3456, out of Picasa uncorrected jpeg 4640 x 3472.

No attempt to make the two look the same, the out of camera jpeg is wishy-washy for other reasons and I haven't bothered to fine tune it just for this example.

The Picasa jpeg has the better sky etc. Never mind the quality, feel the width. Both resized to 560 x 420 for the gif.



Not sure what it proves, but hey, it's fun to see it unbend like this.

There are a number of factors that go on with this rectilinear (curvilinear) stuff. First off, there is the actual spatial content in the X and Y axes of an imaged scene that is projected onto an image-sensor in the first place when these two types of radial lens-system image distortion exist:

Source: http://toothwalker.org/optics/distortion/distortion.gif

Note that in the case of barrel-distortion (center), more spatial content of the imaged scene will be projected onto the active surface of the image-sensor than in the normal case (left) -whereas in the case of pincushion-distortion (right) less spatial content of the imaged scene will be projected onto the active surface of the image-sensor than in the normal case (left).

Then, there also exist matters pertaining to the particular geometric scaling and cropping of the resulting corrected image that is chosen to be presented within the output image-frame.

(Presumably), the choice of the cropping involved (should) equal the maximum possible (vertical, horizontal, and diagonal) amounts of the corrected image that can be displayed within a rectangular output image-frame of some number of pixels arranged in some particular aspect ratio.

What you have been demonstrating relates to the choices made by various in-camera JPG-engines as well as RAW processors and simple RAW converters that do not perform software corrections. The image scaling and cropping that takes place can result in seemingly confusing results.

The important properties to note are that a lens-system having barrel-distortion is able to project optical information relating to a wider and taller (horizontal and vertical, and thus larger diagonal) Field of View than the normal case, and a lens-system having pincushion-distortion projects optical information relating to a narrower and shorter (horizontal and vertical, and thus smaller diagonal) Field of View than the normal case. That is the important thing to understand.

No "machinations" are able to correct optical information that the lens has not projected onto the active surface of the image-sensor (whether than information is geometrically distorted or not).

.

(Presumably), if and when a M43 lens manufacturer specifies the diagonal Angle (Field) of View for a lens-system at some particular Focal Length, they are taking into account the FOV (relative to a zero rectilinear distortion case) that will be gained (in the case of corrected Barrel Distortion) or lost (in the case of corrected Pincushion Distortion) when the optical information that is projected by that lens-system onto a M43 camera body's image-sensor photo-sites - where some (relatively quite small on a percentage basis) number of additional active photo-sites that are usable for imaging surrounding the standard number of photo-sites (4608x3456 for 16 Mpixel 4:3 AR) exist in order to accomplish necessary software corrections for geometric distortions.

 

[Guy Parsons]

Good stuff there DM, I've been reading along with everything, but being a lazy sod I never bother with the mathematics, I just push the button and see what I get.

A brain spasm earlier and in threaded view my Curious Happenings post should have been here but ended up above somewhere next to the legacy lens stuff.

That one shows that Oly Viewer 2 does NOT duplicate what the camera does.

Regards....... Guy

 

[Detail Man]

Guy Parsons wrote:

Good stuff there DM, I've been reading along with everything, but being a lazy sod I never bother with the mathematics, I just push the button and see what I get.

With all that goes on surrounding this stuff, it is understandable that attempting to empirically gather information - by trying this or that processor with various shots having barrel, pincushion distortion, or (as in the case of my LGV 7-14mm at 12mm example) essentially no geometric distortion - can indeed be quite confusing (as well as quite misleading). Proceeed with caution. Too many variables involved.

A brain spasm earlier and in threaded view my Curious Happenings post should have been here but ended up above somewhere next to the legacy lens stuff.

That one shows that Oly Viewer 2 does NOT duplicate what the camera does.

I saw your post. In cases where the differences are so relatively slight, it's hard to conclude much. DxO Optics Pro corrections implement more barrel distortion corrections relative to Panasonic's image-file meta-data directed, in-camera directed corrections (both in-camera, and using Silkypix) with LX3 RW2s at full wide-angle, but (on the other hand) ...

... where it comes to the LGV 7-14mm lens at 12mm on a GH2 (where the rectilinear distortion is nearly zero), DxO Optics Pro corrections implement slightly less barrel-distortion corrections (or perhaps slightly more pincushion-distortion corrections) relative to Panasonic's image-file meta-data recorded, lens-system directed corrections (in the case of the GH2 in-camera JPG).

 

[Steen Bay]

Detail Man wrote:

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

Yes, the "manufacturers must cover the specified FoV in the case of zero rectilinear distortion", but like I said, and like Anders said, and like the WA examples posted by gollywop and Guy Parsons show, then the uncorrected horizontal, vertical and diagonal FoV is quite a bit wider than strictly necessary, so it's perfectly possible to get a rectilinear corrected 4:3 image that has a wider FoV than the OOC JPEG has, a FoV that's considerably wider in all directions than the specified FoV.

 

[Detail Man]

Steen Bay wrote:

Detail Man wrote:

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

Yes, the "manufacturers must cover the specified FoV in the case of zero rectilinear distortion", but like I said, and like Anders said, and like the WA examples posted by gollywop and Guy Parsons show, then the uncorrected horizontal, vertical and diagonal FoV is quite a bit wider than strictly necessary, so it's perfectly possible to get a rectilinear corrected 4:3 image that has a wider FoV than the OOC JPEG has, a FoV that's considerably wider in all directions than the specified FoV.

I do not agree with your statements. Prove your assertion directly using images generated by you. Other than direct evidence that actually demonstrates your assertion, I am not at all interested.

 

[mapgraphs]

Guy Parsons wrote:

Whilst fiddling with camera defaut jpeg and then from Oly Viewer 2 and from Silkypix Pro V5 - all delivering 4608 x 3456 jpegs we see these differences......(resized to 560 x 420)

So even though all three make a corrected 4608 x 3456 jpeg, they are all a slightly different frame. Back to my original thoughts here, "how long is a piece of string?"

Regards...... Guy

Yes. The more I dig into what documentation I can find, the stranger the story. All I come up with is more questions.

 

[gollywop]

Guy Parsons wrote:

Here's the Mk1 Oly 14-42 at 42mm on E-PL5, seems to be about the most pincushion I can muster in the stable.

Out of camera jpeg 4608 x 3456, out of Picasa uncorrected jpeg 4640 x 3472.

No attempt to make the two look the same, the out of camera jpeg is wishy-washy for other reasons and I haven't bothered to fine tune it just for this example.

The Picasa jpeg has the better sky etc. Never mind the quality, feel the width. Both resized to 560 x 420 for the gif.



Not sure what it proves, but hey, it's fun to see it unbend like this.

Regards...... Guy

Guy, do you have the ORF for this image? OOC jpegs don't really give the complete information that's needed. If so, would you be willing to e-mail me a copy?

thanks,

--
gollywop

 

[Anders W]

Detail Man wrote:

Steen Bay wrote:

Detail Man wrote:

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

Yes, the "manufacturers must cover the specified FoV in the case of zero rectilinear distortion", but like I said, and like Anders said, and like the WA examples posted by gollywop and Guy Parsons show, then the uncorrected horizontal, vertical and diagonal FoV is quite a bit wider than strictly necessary, so it's perfectly possible to get a rectilinear corrected 4:3 image that has a wider FoV than the OOC JPEG has, a FoV that's considerably wider in all directions than the specified FoV.

I do not agree with your statements. Prove your assertion directly using images generated by you. Other than direct evidence that actually demonstrates your assertion, I am not at all interested.

Panasonic 7-14 at 7 mm converted by RawDigger (I used the screenshot function ).

After proper correction of geometric distortion:

As processed by LR 4.3 from RAW (the OOC jpeg is cropped identically):

 

[Guy Parsons]

Also it seems that with the E-PL5 and Silkypix Pro V5 I can recover only a tiny insignificant whisker of edge pixels in RAW conversion maximum Trim, not worth the effort.

With E-PL1 it makes a noticeable frame difference and retrieves some of the scene normally chopped off in the default jpeg.

So the Panasonic sensor throws away more edge pixels than the Sony sensor when it makes a jpeg.

Not a worry at all, just a curiosity.

One day when I really have nothing to do I will try and use quick release plates to have the E-PL1 and the E-PL5 see exactly the same frame with say the 50mm macro lens set at infinity (or any lens that has no distortion) and check again what the frame differences are for default camera jpeg and maximum Trim jpeg.

Regards.... Guy (forever wasting time in the quest for knowledge)

 

[Guy Parsons]

gollywop wrote:

Guy, do you have the ORF for this image? OOC jpegs don't really give the complete information that's needed. If so, would you be willing to e-mail me a copy?

Can do, on wrong computer right now but I will PM you a location for that ORF within an hour or so.

Despite my reluctance to waste more time, I might download Rawdiigger and do some snooping myself of what goes on. So far (as mentioned above somewhere) I seem to have found that the E-PL5 can recover less useful edge pixels than the E-PL1 ORFs can manage.

Regards...... Guy

 

[gollywop]

Anders W wrote:

gollywop wrote:

Whoop, Anders, your link to the means for testing angle of view doesn't work. Can you try again?

thanks,

Don't know what went wrong with that. But I corrected the link by editing my previous post. It should work better know (at least it did when I tried) but here it is again just in case:

http://www.panohelp.com/lensfov.html

Well, Anders, it seems my hypothesis (and hope) was wrong. The nominal focal length appears to apply to the uncorrected, not the corrected, image. I did the measurements according to the link you give above. I did them twice, once fairly rough and ready and again much more carefully, but there's no substantive difference; they tell essentially the same story with the 12-35 at 12mm:

The corrected AoV is about 68.67° for an EFL of 26.35

The uncorrected AoV is abut 73.33° for an EFL of 24.23.

So, the 12-35 appears to be a 12mm lens only if you put up with the distortion. It is otherwise a 13 mm lens (at least after ACR does whatever it does -- which, I suppose, is effectively what the OOC jpeg would be also).

--
gollywop

 

[Anders W]

gollywop wrote:

Anders W wrote:

gollywop wrote:

Whoop, Anders, your link to the means for testing angle of view doesn't work. Can you try again?

thanks,

Don't know what went wrong with that. But I corrected the link by editing my previous post. It should work better know (at least it did when I tried) but here it is again just in case:

http://www.panohelp.com/lensfov.html

Well, Anders, it seems my hypothesis (and hope) was wrong. The nominal focal length appears to apply to the uncorrected, not the corrected, image. I did the measurements according to the link you give above. I did them twice, once fairly rough and ready and again much more carefully, but there's no substantive difference; they tell essentially the same story with the 12-35 at 12mm:

The corrected AoV is about 68.67° for an EFL of 26.35

The uncorrected AoV is abut 73.33° for an EFL of 24.23.

So, the 12-35 appears to be a 12mm lens only if you put up with the distortion. It is otherwise a 13 mm lens (at least after ACR does whatever it does -- which, I suppose, is effectively what the OOC jpeg would be also).

Well that's a pity as far as the 12-35 is concerned. But you also have my test of the 20 and Lenstip's test of the 14, where things worked out differently (FoV after correction equals the specs). So my preliminary conclusion, provided your measurements are correct, is that Panasonic has cheated a bit when it comes to the specs of this particular lens rather than it being generally the case that the specs refer to the uncorrected image.

 

[Anders W]

Anders W wrote:

Detail Man wrote:

Steen Bay wrote:

Detail Man wrote:

With respect to some fixed reference aspect-ratio of output images, it seems that camera body manufacturers must cover the specified FOV in the case of zero rectlinear distortion, and we have found that additional FOV can only be realized (in the case of the existence of finite amounts of barrel-distortion) when one is willing to increase the magnitude of the numerical value of the output image aspect-ratio (such that the horizontal dimension of the output image is allowed to increase relative to the vertical dimension).

Yes, the "manufacturers must cover the specified FoV in the case of zero rectilinear distortion", but like I said, and like Anders said, and like the WA examples posted by gollywop and Guy Parsons show, then the uncorrected horizontal, vertical and diagonal FoV is quite a bit wider than strictly necessary, so it's perfectly possible to get a rectilinear corrected 4:3 image that has a wider FoV than the OOC JPEG has, a FoV that's considerably wider in all directions than the specified FoV.

I do not agree with your statements. Prove your assertion directly using images generated by you. Other than direct evidence that actually demonstrates your assertion, I am not at all interested.

Panasonic 7-14 at 7 mm converted by RawDigger (I used the screenshot function ).

After proper correction of geometric distortion:

As processed by LR 4.3 from RAW (the OOC jpeg is cropped identically):

Correction: I didn't check the aspect ratio of the distortion-corrected version of the image converted by means of RawDigger. So here it is again, with the vertical edges cropped so as to conform to the 4:3 aspect ratio (the image is actually 1022x768 so I cropped a tiny bit more than required). The FoV is still clearly wider in all directions than in the LR4/OOC-jpeg version.