Correction of fisheye distortion in LR - useful tool or does it degrade the image?

Image editing software (and build-in firmware of camera bodies) allow for the gradual correction of "fisheye distortion" of images obtained using fisheye lenses. Fisheye lenses are regarded to be valuable for UW WA photography by many. At 180° FOV this distortion is at its maximum and it decreases when, with zoom fisheye lenses, the focal length is reduced. Also WA wetlenses as WWL and WACP, exert to some extend, fisheye distortion.

Since some time I often correct this fisheye distortion (obtained with the Canon 8-15mm and the Tokina 10-17mm fisheye lenses on MFT cameras (Olympus EM1II and EM5II)) in LR, in most cases not 100% but less - and I like it. In a recent thread on the Canon 8-15mm fisheye lens, several peers state, that correction of fisheye distortion in postprocessing is not worth as it ruins the image quality. This makes me thoughtful and I would like to hear the opinion of the others (and maybe see also examples). It would not be the first time that I am excited by some postprocessing gadged, but after using it for some while and hearing the opinion of others, I discover that I went into a dead end...

When discussing the impact of fisheye distortion correction, I believe that one should take into account:

#1.: The enormous AOV of 180° at 8mm (for MFT sensor). A rectilinear lens with comparable AOV should have a focal length of less than 1mm. I guess that in real life the IQ (and also the distortion) would be so bad that thsi is the reason that such a lens does not exist.

#2.: When juding the effect of "defishing" on IQ, this should not be done using regions that are not in focus. E.g. from regions in the corner that are much closer to the photographer as the main subject and therefore not in focus. Stretching out such out of focus regions will lead even to more blurr in the result. This kind of blurr is a different issue as the corner unsharpness that is provided by domeports at wider apertures.

Subsequently I post two examples, how I use this correction tool...

 Wolfgang

example #1.: Table coral. Canon 8-15mm @8mm, ISO 200,  f 10, 1/160s.

Note that the name of the coral is "TABLE"coral and not mushroom coral...

uncorrected image:

80% correction of fisheye distortion (this image I find pleasing and show to others, however a bit rotated to make the table horizontal (the example images are uncropped)):

100% correction ( find this is too much):

Maybe I should add that I use a LR lensprofile that I have generated by myself (Canon 8-15mm fisheye behind Nauticam 140 domeport on EM1II camera). The LR build in profile is for this lens but on FF sensor and I am not sure whether it would work nicely for images obtained with MFT sensor...

example #2.: Inside SS Thistlegorm. I apologize for the backscatter that I did not remove in the example images, but is irrelevant for the comparison. In the image that I normally show to the others, backscatter has been removed in PS, but I am not able to adjust fisheye correction in the final image version...

ISO 200, f 9.0, 1/200s

uncorrected image (For my taste the front wheel and the face of the diver appear, just a little bit, too "bold"):

40% correction (this version, with backscatter removed, I show to others):

100% correction (too much for my taste):

This is exactly what I need to know, I'm between buying a Fisheye EF 8-15 mm (~900€) or new RF 16 mm (~300€).

Since I only like corners straight, I don't know if quality of image in Fisheye 15mm corrected stills much better than RF 16 mm? 

PS: Looking at your images, 100% correction is to avoid! 80% looks much better!

37 minutes ago, stillviking said:

This is exactly what I need to know, I'm between buying a Fisheye EF 8-15 mm (~900€) or new RF 16 mm (~300€).

Since I only like corners straight, I don't know if quality of image in Fisheye 15mm corrected stills much better than RF 16 mm? 

PS: Looking at your images, 100% correction is to avoid! 80% looks much better!

I think that rectilinear WA and fisheye is not exclusive, but they complement each other. It is a matter of personal taste, but since I have the 8-15mm, I seldom use the rectilinear lenses. Few weeks ago in Egypt I could use the Zuiko 10-25mm (similar to 20-50mm in FF and a nice zoom range) for a couple of dives and this confirmed me that I prefer the fisheye. With my wife, who is also a UW-photographer it is almost the opposite, she prefers the rectilinear lenses...

It also should be remembered that my examples are from Canon 8-15mm with Nauticam 140 dome on a small MFT sensor. IQ may well be different when using the lens @15mm (similar to 8mm on MFT sensor) with the same domeport on a FF camera...

Wolfgang

On 1/22/2023 at 2:42 AM, Architeuthis said:

 

#1.: The enormous AOV of 180° at 8mm (for MFT sensor). A rectilinear lens with comparable AOV should have a focal length of less than 1mm. I guess that in real life the IQ (and also the distortion) would be so bad that thsi is the reason that such a lens does not exist.

 

It is actually not possible to have a 180deg image that is also rectilinear.  A rectilinear lens  takes a straight line parallel to the sensor plane and renders it as a straight line in the image.  If a line is parallel to the sensor plane the end of the line will be at infinity at 180 deg.  If you look at the formula for rectilinear field of view :

To get 180 deg field of view focal length must be 0 mm, which you can't calculate (divide by zero) which I believe is mathematically representing what I described above.   

If you look at de-fishing methods you'll see that it needs to crop out the extreme corners.  Here is a link, scroll down to point 4 the example and it will show you the uncropped image.  The wildly distorted corners are cropped out.  

Correcting Fisheye Images | Baeldung on Computer Science

I would suggest you may get a similar effect by just zooming in a little to match a rectilinear field of view, that way you get more of the sensor used to make the image.  Of course this is only possible currently on a crop sensor camera or a fisheye with tele-converter in the case of full frame.  

2 hours ago, ChrisRoss said:

If you look at de-fishing methods you'll see that it needs to crop out the extreme corners.  Here is a link, scroll down to point 4 the example and it will show you the uncropped image.  The wildly distorted corners are cropped out.  

 

Correcting Fisheye Images | Baeldung on Computer Science

I would suggest you may get a similar effect by just zooming in a little to match a rectilinear field of view, that way you get more of the sensor used to make the image.  Of course this is only possible currently on a crop sensor camera or a fisheye with tele-converter in the case of full frame.  

Interesting link...

Optically zooming in has the same effect on the distortion of straight lines as has blunt digital cropping in post-processing (of course you loose pixels when you crop in LR, while you use the full resolution of the sensor when zooming in): by showing a smaller section of the image, the relative amount of distortion becomes smaller, but it is still a fisheye image (this is why people say the fisheye distortion on, e.g. WWL-1 with 130° FOV is smaller compared to a "real" fisheye (they mean a 180° fisheye with the term "real" fisheye)...

Digital correction of fisheye distortion is, however, a different process. One may say that de-fishing is also some kind of cropping, but the cropping is selective in this case: imagine the image is a rectangular piece of pastry and you stretch it out by pulling on the corners. Then you cut out from the now distorted pastry, that is no longer a rectangle, a straight rectangle (with the maximum area possible) - this is what de-fishing is...

Wolfgang

36 minutes ago, Architeuthis said:

Interesting link...

Optically zooming in has the same effect on the distortion of straight lines as has blunt digital cropping in post-processing (of course you loose pixels when you crop in LR, while you use the full resolution of the sensor when zooming in): by showing a smaller section of the image, the relative amount of distortion becomes smaller, but it is still a fisheye image (this is why people say the fisheye distortion on, e.g. WWL-1 with 130° FOV is smaller compared to a "real" fisheye (they mean a 180° fisheye with the term "real" fisheye)...

Digital correction of fisheye distortion is, however, a different process. One may say that de-fishing is also some kind of cropping, but the cropping is selective in this case: imagine the image is a rectangular piece of pastry and you stretch it out by pulling on the corners. Then you cut out from the now distorted pastry, that is no longer a rectangle, a straight rectangle (with the maximum area possible) - this is what de-fishing is...

 

Wolfgang

 

In the case of the Canon 8-15mm the lens has a horizontal field of view of 142 degrees and a vertical field of view of 92 degrees

This means the lens would return a 2.8:1 aspect ratio. the corners would be cropped they cannot be rebuilt

As @Alex_Mustard said you compose the shot as you see it in the viewfinder you don't think how it will work once is defished so I never defish any shots with fisheye nor zoom fisheye

If you don't like distortion you get a nice rectilinear lens. A 12mm will give you 90 degrees vertical but only 112 degrees horizontal as it preserves the aspect ratio

Comparing fisheye lenses to rectilinear lenses or defishing fisheye lenses will not make the images compareble

And this is why all the discussions about corner sharpness are futile as the field of view can never be matched on all axis.

The edges of the defished shots from Wolfgang look worse than what a rectilinear lens would produce...

34 minutes ago, Architeuthis said:

Digital correction of fisheye distortion is, however, a different process. One may say that de-fishing is also some kind of cropping, but the cropping is selective in this case: imagine the image is a rectangular piece of pastry and you stretch it out by pulling on the corners. Then you cut out from the now distorted pastry, that is no longer a rectangle, a straight rectangle (with the maximum area possible) - this is what de-fishing is...

To illustrate this, defishing a fisheye shot, using a picture in my catalog (manual distortion correction pulled at +100), from a 15mm (on Full Frame).

Edit

Original fisheye shot

Defisheyed shot (manual distortion correction set to +100, rather extreme but it illustrates well what Wolfgang had mentioned)

4 hours ago, waterpixel said:

To illustrate this, defishing a fisheye shot, using a picture in my catalog (manual distortion correction pulled at +100), from a 15mm (on Full Frame).

 

Thanks for posting, I think it serves to re-inforce the point, the original looks great and only subtle differences in the de-fished versions so I would suggest use the fisheye for subjects like this and if you have a wreck to shoot where the fisheye effect is more obvious or other subjects like more distant sharks etc, use a rectilinear.

Bit late to this party but:

I've been humming and hahhing for ages about whether to get a rectilinear for my D500. Since moving back to DX the Tokina 10-18 has been my go-to wide-angle lens. It is very rare indeed that, for me, the fisheye distortion is noticeable and unpleasant. 

Looking at Wolfgang's table coral examples, it's pretty hard to tell the first one is a fisheye shot. Does it need adjusting? It's in the eye of the beholder. Maybe the Thistlegorm shot exaggerates the perspective slightly. But some would argue that it increases the dramatic effect.

As Chris suggests, and the conclusion I have come to, that adding a rectilinear is helpful if you want to shoot subjecs that are a bit more distant (as Chris says, sharks, whales etc) or ,interesting point made by Massimo,  if its fast moving subjects that might well end up near the edge of the image and which will need cropping (though this then gets into the soft edge issue)

For those reasons I decided recently to buy the Nikkor 10-24 - essentially to have a lens that is better for those bigger, slightly more distant subjects. Lots of sharks here in Sint Maarten! I hope to get it in the water next week.

39 minutes ago, TimG said:

Bit late to this party but:

I've been humming and hahhing for ages about whether to get a rectilinear for my D500. Since moving back to DX the Tokina 10-18 has been my go-to wide-angle lens. It is very rare indeed that, for me, the fisheye distortion is noticeable and unpleasant. 

Looking at Wolfgang's table coral examples, it's pretty hard to tell the first one is a fisheye shot. Does it need adjusting? It's in the eye of the beholder. Maybe the Thistlegorm shot exaggerates the perspective slightly. But some would argue that it increases the dramatic effect.

As Chris suggests, and the conclusion I have come to, that adding a rectilinear is helpful if you want to shoot subjecs that are a bit more distant (as Chris says, sharks, whales etc) or ,interesting point made by Massimo,  if its fast moving subjects that might well end up near the edge of the image and which will need cropping (though this then gets into the soft edge issue)

For those reasons I decided recently to buy the Nikkor 10-24 - essentially to have a lens that is better for those bigger, slightly more distant subjects. Lots of sharks here in Sint Maarten! I hope to get it in the water next week.

Tim I was referring to topside shots like birds in flight not underwater when talking about crop

For me underwater rectilinear lenses have three purposes 

1. Divers or wrecks where I can clearly see distortion (not all wrecks are like that)

2. Split shoots where you can see the horizon line

3. Blue water shots when a fisheye makes thing look tiny

In general I have full cover from 20 degrees to 180 degrees with my underwater kit

Macro

Standard lens (more useful for video) up to 85 degrees

Wide rectilinear up to 105 diagonal (not more as the edges suffer and the image looks weird)

Zoom fisheye 90 to 180

There is no overlap or limited overlap between the fisheye and the rectilinear. Yes you can shoot WACP etc etc fully zoomed but actually they do not look as nice

Field of view increases as you get close so with a rectilinear lens not having it too wide helps as it avoids the edges issues as you need to step back

I am sure most would prioritize a fisheye and a zoom fisheye (WACP or Tokina) style before a rectilinear lens however if you shoot topside you will already have a rectilinear lens

In addition some people like Dr Mustard shoot a fisheye with a 230mm dome so the bulk is already there

I would also argue that with the 3.9 kg weight and 20 cm diameter of the WACP the space saving is zero is all about quality and having that zoom fisheye that full frame users do not have

1 hour ago, Interceptor121 said:

Tim I was referring to topside shots like birds in flight not underwater when talking about crop

For me underwater rectilinear lenses have three purposes 

1. Divers or wrecks where I can clearly see distortion (not all wrecks are like that)

2. Split shoots where you can see the horizon line

3. Blue water shots when a fisheye makes thing look tiny

All looks good to me, Massimo