From my understanding, a fisheye lens usually has a coverage of 180 degrees or close to it. Given the extreme coverage and the availability of software that can turn fisheye images back to a rectilinear perspective, why are fisheye lenses not more popular, considering their usually much cheaper price compared to rectilinear wide angle lenses?
10 Answers
When you correct the distortion in an image from a fisheye lens, you get undesirable side-effects.
You lose a lot of LOT of diagonal angle of view from cropping, to get a rectangular image out of it. See the below example of a rectilinear conversion (yellow indicates the largest usable rectangular area after fisheye to rectilinear conversion). So after correction, you have lost some image information.
You lose a LOT of resolution on the corners of the frame. For an example, look at the corners of the first example image (above). They are really blurry the further you get to the corners.
Rectilinear doesn't mean no distortion. Simply due to the huge angle of view, you'll still get things looking very stretched at the edges, even though it's technically what you'd get from a rectilinear lens that had the same angle of view. See this second example of a rectilinear conversion (below), which has already been cropped to the largest usable rectangle. The people on the far left and right look stretched, even though that's how they actually look, if you were able to photograph them with a really wide rectilinear lens. You can verify that it's a rectilinear projection because the straight lines are all straight, not curved.
In reality the photographer would have been very close to this group of people and using a fish-eye lens. At least the people on the ends wouldn't have looked so wide in the original fisheye!
Also, even in the previous image I showed you of the buildings, notice how exaggerated the divergence of vertical lines is, since the camera is slightly angled upwards.
-
\$\begingroup\$ Thanks for your answer! I changed your answer to the accepted one because of the great explanation and example images you provided. \$\endgroup\$ Feb 4, 2011 at 20:05
-
\$\begingroup\$ Can the horizontal field of view of the largest usable rectangle be computed from the field of view (or focal length) of the fisheye? \$\endgroup\$– mattdmMar 11, 2011 at 14:22
-
\$\begingroup\$ Probably, but don't look at me for the formula ;) I think it would depend not only on the field of view of the fisheye and the aspect ratio, but also on just how distorted the fisheye is. I doubt that one 10.5mm fisheye would have the same distortion as another 10.5mm fisheye. \$\endgroup\$ Mar 15, 2011 at 4:00
-
\$\begingroup\$ doesn't your third point - that "A rectilinear image of such a wide angle of view makes things look very stretched at the edges" - apply equally to both a corrected fisheye image, and to one taken with a rectlinear ultrawide lens? \$\endgroup\$– bacarApr 11, 2011 at 16:34
-
\$\begingroup\$ @bacar, yes, that was what I was trying to say. At that extreme angle of view, a rectilinear image will look stretched at the edges whether it's a corrected fisheye or an equivalent hypothetical rectilinear lens. \$\endgroup\$ Apr 13, 2011 at 0:15
You lose a ton of resolution when you essentially crop a small portion of your image. Also because of the curvature of the image retained resolution will not be even across the image, which can wreak havoc with apparent sharpness in a print (or even just a web image).
Lastly it's a lot of post processing work you'd have to do for every image you care about. Life is SO much easier if you get things right in-camera first, or as close as you possibly can.
First, who said Fisheye lenses are cheaper than wides? Currently the old EF 15mm fisheye is sold for ~$650. The EF-S 10-22mm is slightly more than $700 now. You get all the advantages that @Mike listed from the UW, plus a zoom range. The EF 8-15mm Fisheye was just announced and I can't even find a price quote, but given it is an L lens, expect price >>$1,000.
Then, the amount of distortion is such that the correction in post will necessarily result mediocre rectilinear results. The outer areas are just too compressed (spatially) to be expanded while effectively retaining all the fine detail (the sensor's resolution is not infinite!).
Last - who wants to have to deal with this software process every time a wide angle image is taken? It becomes tedious.
-
1\$\begingroup\$ I definitely agree with the last comment but I'll say it a bit differently - it's better to get it as close as possible in camera as it creates less computer work later to "fix" it. \$\endgroup\$– MikeFeb 2, 2011 at 5:39
-
\$\begingroup\$ "Tedious" is only part of it when you're on the clock or working on contract and have to deliver some images. Having to waste even one minute per image when processing several hundred becomes unbearable. \$\endgroup\$– GregFeb 2, 2011 at 6:03
-
-
\$\begingroup\$ @Greg - agreed. I am speaking from a hobbyist point of view... \$\endgroup\$– ysapFeb 2, 2011 at 11:37
-
2\$\begingroup\$ @Daniel T. - The difference is too small to justify all the hustle. In a good day you could get the 10-22 for about $650 (like what I got it for, but this was really a special occasion) \$\endgroup\$– ysapFeb 2, 2011 at 11:38
A few thoughts that come to mind:
- the distortion in fisheyes is probably more difficult to correct due to the extreme distortion present;
- they may be more susceptible to flare (wider lenses usuallly are more susceptible than more telephoto ones);
- the optics may be generally of lower quality than ultra wides;
- the front element will bulge out further so will be more susceptible to damage;
- good luck using filters with them.
Maybe you could go looking at some lens reviews. photozone.de has some reviews of fisheye lenses among their large number of reviews they've done.
- it requires post-processing (extra time and work required for each picture)
- it requires post-processing (can't see what the final result will look like when framing the scene through the viewfinder or on the LCD after taking the photo)
Some people do. Here's a blog post from 2009 about using the Zenitar 16mm fisheye lens and the ways the images it produces can be manipulated. Lots of pictures.
-
\$\begingroup\$ Thanks for the link. Despite what is mentioned above with loss of quality the images on that page are pretty spectacular. I kind of like the distortion. \$\endgroup\$ Mar 11, 2011 at 22:03
-
\$\begingroup\$ Great cheap lens. Also note that its a fisheye on full-frame so resolution loss is lower than normal. \$\endgroup\$ Apr 7, 2011 at 0:14
-
\$\begingroup\$ For the loss of quality, look at the corners. Such as here and here The low resolution of the uploaded photos makes closer examination impossible but you can still see the level of interpolation that was needed to correct that distortion in right in the corners. \$\endgroup\$ Jun 16, 2020 at 4:53
I use a pocket-size fisheye adapter on a superzoom camera as my wide-angle adapter. This affords me a seamless zoom range from 9mm-36mm (35mm-equivalent) focal-lengths on my superzoom cameras. On one camera it even gives me an effective widest aperture of f/2.0 throughout that whole range. The cameras' own zoom lenses then providing focal-lengths beyond that range. I found one fisheye adapter-lens for less than $100 that is nearly 100% chromatic-aberration free, more-so than even dedicated Nikkor fisheye lenses for D/SLR cameras when I've tested the resulting images against each other. When the superzoom camera's zoom+fisheye adapter is set at 18mm fl, then I get a full-frame corner-to-corner wide angle image. Below that then it starts to vignette until I get a full circle fisheye image at 9mm.
Apart from the issues everyone mentioned above:
- Not everything looks good in fisheye. It looks good once in a while but you can't do too much with the extreme perspective distortion that comes with using a fish eye lens.
- Color fringing is another issue which keeps me away from fish eyes.
-
\$\begingroup\$ colour fringing isn't a product of using a fisheye, it's a product of using a poorly designed or constructed lens. \$\endgroup\$– jwentingMar 11, 2011 at 6:57
-
\$\begingroup\$ Don't know why this got a -1, I bumped it back to 0. The original question was highly subjective. Despite all the mentioned downsides to converting fish eye shots back to rectilinear I happen to like the results of the shots. The extreme distortion adds a nice flair to the shots. There are many awesome examples out there. \$\endgroup\$ Mar 12, 2011 at 16:23
Focal length not only affects field of view, but perspective and depth of field. Even a 50mm full frame shot will be different than a 32mm APS-C shot with the same framing and composition.
- Depth of field is different
- Perspective and warping is different
- Pixel detail is different
For an extreme situation such as that suggestion in the question, you would end up with a significantly worse image than if you simply used the correct focal length.
Having only recently acquired a fisheye I'm having fun with it but I don't see it as a replacement for a regular wideangle lens. I do see it as an alternative to the very heavy 12-24 zoom I'd otherwise buy (I've chronic back problems, the lighter weight was the deal maker for me to buy a 10-17 fisheye rather than a 12-24), but not a direct replacement (I already own a 20mm f/2.8 so wouldn't need the full 12-24 range anyway).
The Tokina 10-17 has very profound curving at 10mm, reduced to barely visible curving at 17mm. It serves its purpose as an ultra wide lens to give me reach beyond the 20mm I was previously limited to, while offering at the same time a nice toy to experiment and get strange visuals if I want to.
And no, they're not cheap. The 10-17 cost me the same the 12-24 would have from the same product range (Tokina ATX-Pro). I believe Nikon and others have similar price simularities.