Fun with Fisheyes

[dougsmit]

Fun with Fisheyes

My Experiences with Low Priced Fisheye Lenses

Fisheye lenses provide an alternate way of looking at the world in wideangle. Regular, "rectilinear", lenses are "corrected" to keep straight lines straight much in the same way as seen in Mercator projection maps of the world. Straight lines are straight but at the cost of severe distortion of the sizes of objects nearing the corners. In a map, this leads people to believe that Greenland is as large as Africa. The relative merits of various map projections is a bit of a political hot potato so I will drop that subject with this mention. Unfortunately it will be more difficult to remain neutral on the question of ultrawide images. In rectilinear wide angle photography, round objects tend to become ovals pointing toward the corners. The "problem" exists at all focal lengths but only becomes an issue in the realm of the wide angle. It becomes more extreme as the proportion between focal length and sensor size decreases. This essay only addresses use of Canon 1.6x crop factor DSLR's where the question arises as we go below 24mm focal length and becomes important below 17mm. Which system is better? That is a question each photographer must address for each use and will appear again and again as this essay continues. I have two fisheye lenses: The 8mm f/3.5 Peleng and the 16mm f/2.8 Zenitar. Both are by far the lowest cost lenses of their respective types. Opinions about the fisheye way of seeing presented here might apply to other brands (Sigma makes an 8mm and a 15mm; Canon makes a 15mm) but I have no experience with these lenses and can not offer a valid opinion on whether they are worth the considerable extra cost required over my two choices. This page considers the Peleng 8mm and the Zenitar 16mm.

16mm f/2.8 Zenitar

The 16mm f/2.8 MC Zenitar is a full frame fisheye delivering a full 180 degree diagonal image on a film camera. However, much of this extreme coverage is lost when used on a 1.6x crop digital. In fact, many of the images need to be examined closely to see that they have the fisheye distortion. The mount is heavy and well built as expected of a Russian lens. The diaphragm is manual lacking the pre-set feature found on the Peleng. This makes focus at smaller apertures more difficult. Still, focus accuracy is significant especially if working at closer range or wide aperture. The front of the lens has a two piece petal hood and no filter threads. The strongest feature of the lens is its price - often $125 to $150 new on eBay. Following the lens name on the barrel will be a letter indicating the mount originally on the lens (M,K, N for M42, Pentax K and Nikon) but many have been modified to have a Canon mount without changing this letter so my example still reads 'M'. I recommend buying the factory M42 version and using an adapter for M42/EOS. My direct to EOS example and many others, so I hear, were converted with litle regard to focus to infinity. I had to loosen three screws under the rubber of the focus ring and correct the infinity position. Now my lenshood is slightly crooked. That would be three more even smaller screws and not worth the effort since I am using the lens on a 1.6x camera and the hood is too short to matter.

Using any lens on a 1.6x crop camera results in the lens appearing to be less wide angle than it did when used on a 35mm film camera or full frame DSLR. If fact, the difference is the factor 1.6x. On a crop camera we need a 10mm lens to simulate the angle of view of a 16mm lens. Since most lenses are still made to cover film, we are faced with paying for complicated designs that can cover a large area which we do not have in our cameras. The 1.6x crop factor applies to fisheye lenses but the difference is actually a bit more than most people expect. Fisheye distortion compresses the edges of the scene more the further you get from the center. When you crop off the outside areas of the full frame, the part that is lost is the most compressed and the part that shows the most of the fisheye effect. If you dislike the effect, that is good but, if you are happy with your fisheye on film, the 1.6x crop version may be disturbingly ordinary. This certainly happens on the 8mm Peleng but it was so wide that there is still a lot of fisheye effect left after the crop and much of the loss was black ends of the rectangular frame left untouched by the circular image on film. The 16mm Zenitar suffers a greater proportional loss in this regard. Each of us needs to look at images and consider what we are wanting to do before selecting one foca length over the other. My opinion: Fisheye fans will want both lenses while those who don't like one will not like the other. If you can not learn to love the look of the fisheye view of the world, I suggest you not buy one. I love mine - both of them.

Many people take distortion much more seriously than I. My first example photo demonstrates this point. Yes, the tree in the background is curved. Are you sure that is a factor of the lens and not the way it grew? The distortion present in the 1.6x crop part of the image is small enough that many images can accept it without looking fishy. Certainly that is not the case with architectural images but that is the situation for which software like PTLens was made. I prefer leaving well enough alone on shots where the corner spread of a rectilinear lens might be as objectionable as the fisheye effect. This image was taken from about arm's length. I know this from the fact that my hand appears at the bottom of the scene. The compact lens allows easy one handed camera operation when you need the other for better purposes (like feeding the subject). I usually try to shoot the Zenitar at f/8 or smaller aperture because it becomes quite sharp in this range. Stopping down makes focus by scale (guess it and set it) as likely to be correct as trying to find a sharp point through the finder. Focus was pre set where the tongue would be and the shot was taken when the moment and spacing came together. Since the shot was taken in bright light it was possible to stop down to get considerable depth of field. This was probably f/16 but I really do not know since the EXIF data did not record it and neither did I. I really wish I had bought the M42 version of this lens so I could add a Focus Confirmation adapter. While depth of field is considerable, I still lose more shots to focus error than anything else. Perhaps that is because I tend to push the envelope with this lens and try to take shots beyond those usually intended for this type of equipment. I love using the lens up close. The focus scale ends at 0.3 meters (about a foot) which is not close enough for many shots I have taken. The screw mount version would allow partly unscrewing the lens to add a couple millimeters to the focus travel. There are other answers but I still wish I had the M42 version (anyone want to trade?).

The museum view above (left) demonstrates an image that does show the fisheye distortion and could have been corrected in software. Just to prove I'm not too lazy, I used PTLens to correct it and placed the results on the right. This correction causes some changes in cropping and proportions which can be controlled in the software but this will serve to give the idea of the effect. The original shows bending of the straight doorframe while the corrected version distorts the circle of light at the top and affects the shape of Caligula's head (not for the better, in my opinion). Of the two, I prefer the fisheye. The Zenitar works well for museum shots in close quarters. The room in which this shot was taken was not large. Lighting was not good and both flash and tripods were forbidden. The choices I had that day were the kit lens at 18mm and the Zenitar at 16mm. Because of the fisheye distortion, you actually get quite a bit more coverage with a 16mm fisheye than an 18mm rectilinear. The image was shot stopped down a bit to increase depth of field and sharpness (f/4, I believe but again I made no record). The ease of hand holding offered by this lens played a part in the success of this image. Perhaps I should have stopped down even more. Handholding a 16mm lens at 1/30th second is no problem.

Update: I became aware of another option for correction of fisheye images that deserves mention. RectFish improves the mapping of Peleng and Zenitar images (I have not tried it on other originals) using a routine somewhat different from PTLens. It does a wonderful job straightening things out without adding the horrible flying corners of standard rectilinear adjustment (which the program offers as a bonus option as well). It crops a minimal amount of the image information in the process. On the down side, currently the program is only adjustable by inputting camera+lens combinations so there is little to be done in the way offine tuning. My limited experience with the program suggests that the results are good and I can recommend it to anyone willing to pay the price for defishing. RectFish appears to be a work in progress (notably the promise for fine tuning and rotational controls) so I hope it will develop into something really great with future updates. Its author appears to be on the right path toward a great tool for those of us who work with fisheyes.

Our next example (Waterfall at Gauley Bridge, WV) demonstrates what I consider to be the perfect situation for the use of a fisheye. You might ask why the image, then, is not perfect? The answer is that I took it. It has nothing to do with it being a fisheye shot. There is every bit as much distortion and corner compression in this image as there is in any other image on this page but the viewer lacks any points of reference (straight lines) that shout out that the fisheye was used. A rectilinear wide angle also would be able to make a nice photo here but it would most likely leave artifacts of the corners flaring out and look less natural than this view. Landscapes often can make good use of a fisheye since they can give a wide view in a restricted space. This scene would be hard to take with a less wide angle lens (I was standing on a rock in the middle of the water - backing up was not recommended). This cramped scene lacking straight reference lines is a situation tailor made for the Zenitar 16mm. This image was taken at small aperture (probably f/16?) to allow front to back focus. The nearest points were barely a foot away. The depth of field scale on the lens is based on the needs for a film camera so it is necessary to focus a bit more exactly than it would suggest. I allow two stops difference when using the lens for images that will be printed at 12x18 inches (the largest I do). That means I shoot at f/8 and keep my distances within the f/4 marks on the lens. Shooting, as here, at f/16 means that I was willing to sacrifice some sharpness to diffraction in order to get the desired depth of field.

When I started to write this page I realized that I did not have an image to illustrate use of the Zenitar in low light situations. As mentioned earlier, the lens is pretty sharp at f/8. It is less so wide open. However the unsharpness seems to be a soft halo (spherical abheration?) surrounding a centrally sharp image rather than an even blur. This responds reasonably well to Unsharp Mask in postprocessing. To correct the lack of image to illustrate this point, I took the lens on our lunch trip and photographed my grandson playing a game while waiting for our food. This time I did take note that the aperture was f/2.8 (wide open). Certainly the quality is lower than some might approve but the wide open lens is useable for noncritical work if you can find a way to be sure the focus is right on. Focus in this light was very uncertain to me and may be just a bit behind the proper plane. Had I not been intent on shooting something wide open, I probably would have boosted the ISO a stop or two and stopped the lens down to improve the sharpness on both counts. Some unsharpness here may attributed to the slow shutter speed which failed to stop his hand as it moved a golf tee from hole to hole. The shot does show some curvature but nothing I consider needing correction. The inset shows a 100% crop section of the eye from this image. Obviously the lens is not 'L' quality but it is usable for small prints or emergencies. Try to use the lens stopped down whenever possible. Even if the lens were sharper wide open, the difficulty of manual focus would drive me to stopping down at least a stop or two.

Close up photos (defined as things not close enough to be called true 'macro' but still showing something up close and personal") provide another interesting use of the fisheye. The short focal length of a fisheye means that these close ups are made almost on top of the subject with almost no working distance when compared to the usual 50mm to 180mm macro lenses. Most people will tell you that wide angle lenses are not suitable for 'macro' and they are correct about 90% of the time. However, there are situations where the perspective provided by a wide lens can open up new interest in some images. The example here is a very spidery flower (Hymenocallis) with long, slender petals and stamen that compress unnaturally when shot with a standard, long focus macro. This same flower is shown with 102mm and 8mm renditions in my Hymenocallis Gallery. Those interested might enjoy comparing the 'look' of the stamens and pistels in particular. There is another downside to wide angle macro. Many photographers highly value the ability to separate the sharp subject from the out of focus background. When focussed on three inches, even a fisheye lens renders infinity out of focus. Unfortunately it is not very far out of focus so the details of the background lack the creamy bokeh so valued by most users of a DSLR. Even at the smallest aperture, one does not get perfect sharpness from one inch to infinity so the average fisheye macro result is caught in the middle ground between sharp and soft that is not appreciated by proponents of universal sharpness or selective focus. The Zenitar will never be my prime macro lens but getting close is fun.

Since the Zenitar only focuses to 0.3 meters it is not immediately suitable for macro use. Adding the smallest available extention tube (12mm) works but leaves a huge gap between the normal near and the tube's far distance. No one makes the needed 6mm extension tube for Canon. My answer was to use an old 49mm +6 diopter close up lens left over from my point and shoot camera. The Zenitar front is 62mm diameter but has no threads to hold a filter. I bought a 62-49 adapter ring and taped it in place. Due to the 1.6x crop factor there was no vignetting even with this small filter. It would be better to use a larger filter but this started as an experiment with available equipment. Below is the first image made in this manner. The nearest point of the subject is two inches from the lens. The perspective here is considerably different from a normal view of these toys. Other strength diopters will allow other subject distances.

In summary, the Zenitar 16mm Fisheye is far from perfect but can provide a fun experience and decent image quality with a little effort (and stopping down a couple stops). Certainly I would prefer a better lens with autofocus and, especially, auto diaphragm but the extremely low price of the Zenitar makes it a deal hard to pass up. I recommend its purchase to anyone not immediately comfortable with the price of the Canon or Sigma 15mm lenses.

---

8mm f/3.5 Peleng

The MC Peleng 8mm f/3.5 is made in Belarus and sold primarily by mail either from eBay or other Internet sources. Usually it is priced between $250 and $300 US. My lens came packaged with an M42 (Pentax screw) mount on the lens and a manual Nikon F mount (requiring a screwdriver to install) in the box. It also came with a Japanese M42/EOS adapter which was added by the seller. The diaphragm on the lens is sometimes called "manual" but it is actually what used to be termed "Pre-set". One ring allows setting the desired aperture while a second requires turning from "Unlock" to "Lock" to actually stop the lens down to the preset setting. This makes it considerably easier to open up the lens for focus and stopping down to working aperture without removing the camera from the eye. The mount is metal and has a solid feel some might term more "clunky" than "refined". I consider it well built compared to consumer grade Canon lenses but it is no 'L'. Infinity is on the right end of the scale making the lens match Canon and Sigma lenses and run opposite Nikon and Pentax. The opposite end of the scale bears an M (macro?) which is just slightly closer than the last marking of .22 meters. "MC" in the name denotes multicoating.

The package (often, and correctly, demeaned in reviews as crude) also contains an imitation leather case and a set of colored filters which can be used in place of the clear one mounted on the rear of the lens for Black and White film photography. With digital camera (all I use these days), these are unnecessary so mine remain wrapped as received. One or the other of the rear mount filters must be used at all times or the lens will lose the ability to focus to infinity. The lens comes with a metal lenscap to protect the protruding front element. This is held in place (in theory) by friction and a way too thin ring of cloth that guarantees the cap will fall off regularly. I replaced this worthless material with a strip of black gaffer's tape and the cap fits perfectly. Your answer may differ but you will need to do something about that lens cap or it will be lost before you learn to spell "Peleng". While I'm sure I could drone on and on about other details, I'll skip ahead to the pictures and use examples to illustrate other points that seem worth mentioning. Examples here were taken with either my old Canon Digital Rebel (300D) or my new (newer to me than the Peleng) Canon 30D. I have no experience with the lens on other cameras (film, Nikon or otherwise) so don't ask. Rumors say the rear of the lens (or more exactly: the filter) strikes the mirror on some Nikon products. I once heard that someone had a problem with this lack of clearance on a Canon 5D but others reported no problem. I do not know the answer here but would warn all that the required (as part of the lens design!) filter must be completely screwed into place or it might stick back into the camera far enough to touch the mirror. There is no problem with the two cameras I have tried and I have heard nothing about any problems with other Canon models.

Part of me thinks I should show direct from camera images and there will be at least one somewhere down below but my experience with the Peleng suggests that I need to postprocess and crop to make images that look good enough that I am willing to show them to you. Cropping is needed to remove the ugly (my opinion) black corners and ragged edges of the image circle as seen on 1.6x crop cameras. The lens produces an almost circular image on a film camera or a no crop factor digital. The circle is clipped slightly at top an bottom (why they did not make it a 7mm, I can not say). On a 1.6x camera, the four corners are cut rather like you get using a telephoto lens hood on a wide angle lens. Making a rectangular image, therefore, requires cropping to the center section or expanding the corners by use of a defishing program. I will cover my experiences with PTLens in this regard later but it will suffice to say that there are several ways of correcting this situation and each of them has its plusses and minuses. My sandbox photo shows distorted trees that were not cropped out but my grandson (centrally located) looks pretty normal except for the enlargement of his legs which were slightly closer than the rest. The sandbox was round so it takes will with a lens that likes round things. At the top we see a small tree almost knocked down by a larger tree during Hurricane Isabel. It, and some other background trees, are bent in an unnatural way but trees are not always perfectly straight so you might have trouble telling which distortions belong to nature and which to the lens. This was taken from about three feet (just outside the sandbox) and was cropped to straighten and eliminate excess woods. Focus was done by a combination of eye and scale guessing. There is a lot of depth of field with this lens so subjects beyond this distance are pretty much considered at infinity unless you are shooting wide open (this is f/8).

Another way to fill out the corners is to at least partially remove the fisheye distortion and stretch the image into those black spaces. My self portrait above demonstrates this. There is still some curve to the window frame and some other straight lines but I prefer to accept this instead of suffering other distortions introduced by the defishing process. I use (and recommend) PTLens as a plug in to Photoshop Elements. There are other brands which you are perfectly free to explore. This image was taken with the camera tethered to the computer. On the screen you see the image previous to this one since there is no live preview on a DSLR.

Some people prefer rectilinear images to the point that they would software correct every shot. I disagree. Above we see the same photo with and without PTLens correction. On the left we see the out-of-camera fisheye result complete with rounded window frames. On the right is the same image corrected in PTLens so the window frames look better but the child's head is now stretched into the corner. This result is very much like what we would get with a 10mm rectilinear lens from the same position. Opinions vary; I'll take the fisheye version.

Update: I became aware of another option for correction of fisheye images that deserves mention. RectFish improves the mapping of Peleng and Zenitar images (I have not tried it on other originals) using a routine somewhat different from PTLens. It does a wonderful job straightening things out without adding the horrible flying corners of standard rectilinear adjustment (which the program offers as a bonus option as well). It crops a minimal amount of the image information in the process. On the down side, currently the program is only adjustable by inputting camera+lens combinations so there is little to be done in the way offine tuning. My limited experience with the program suggests that the results are good and I can recommend it to anyone willing to pay the price for defishing. RectFish appears to be a work in progress (notably the promise for fine tuning and rotational controls) so I hope it will develop into something really great with future updates. Its author appears to be on the right path toward a great tool for those of us who work with fisheyes.

The above example was corrected with RectFish, otherwise uncropped or manipulated. With PTLens, I had, and used, the option of partially correcting the image but RectFish gives me what it considers proper. I could cause more or less correction by lying to the software regarding the camera and lens used but fine tuning controls will have to wait for another update from the one current when this was written. As it is, RectFish does a fine job of filling in those black corners without requiring excessive cropping or stretching.

The Peleng diaphragm has nine blades producing an interesting 18 pointed star if a strong light source is included in the frame. Nine blades on an 8mm focal length means that each blade is extremely small. It would require extremely tight tolerances to produce a mechanism so small that would give absolutely even results. This was lacking in the Peleng. Add to this the fact that there will be distortion if the light source is anywhere but dead center in the image and the star effect can be shaped a bit oddly. Flare from the sun in the image was only moderate and not as serious as some results when the sun was just outside the scene. There is no provision for a lenshood on the lens but I have considered trying to fabricate something appropriate considering the image reduction that comes with the 1.6x crop camera format. So wide an angle of view makes for a challenge to keep the sun out of the frame if it is anywhere except directly at your back. In that case, the problem becomes keeping your shadow out of the picture. Flash photos with the fisheye are possible as long as you use bounce and can find a bounce surface that is not in the scene. That can be troublesome as well. I have done some images with flash fill using the bubble bag or other diffusers discussed on my flash diffusers page but care must be taken that no part of the light from the flash goes forward an strikes the glass. The usual answer is to rotate the flash head and tilt it slightly backwards. The churchyard image (St. Peter's Episcopal Church, Lewes, Delaware) was partially corrected in PTLens to straighten out some curves so the round sun became elliptical. Some healing brush work was required to remove small flare spots. The odd look on the lawn is due to it not actually being grass but a succulent ground cover. I have found that many Peleng images require more attention in postprocessing than "normal" photos. It is not a toy for those purists who believe that postprocessing is a sin. I doubt that this is greatly different with any 8mm lens so it should not be considered a fault unique to Peleng.

As promised I will provide an uncropped and unstretched shot (above left) and, just to fill space, will show (right) a 100% crop of a portion of that same image. Most obvious is the set of four black corners. At the top is a bright blue rim of flare on the light side of the image while the bottom shows a wider (but less obvious) band of lowered contrast flare. The severity and placement of these faults varies with the image but one or the other appears in most images. The blue rim is less a problem since it is narrow enough that it crops out easily. The low contrast band is a bit harder to crop sometimes but can be locally enhanced with careful use of selection and contrast controls. This is not my sharpest Peleng shot but will serve to illustrate that the lens is somewhat short of 'L' glass quality. I believe it is in the general area of the kit lens image quality (possibly a bit better). Certainly it makes fine 4x6 inch prints and 12x18's are good as long as you view them from a respectful distance for a print that size. Remember that a 100% crop of an image this size simulates a 32x48 inch print (assuming 72 dpi on the monitor) so fair viewing of a crop requires standing across the room. For the record, this image shows a blue stain on the boy's face and hand. This is not a photographic fault but comes from rubbing an inked hand stamp (for being a good boy at pre-school) across your mouth.

I really dislike the black corners and the flare edges. They make it difficult to crop an image and maintain the proportions desired without cutting out something significant. I tried using a 1.4x Tamron converter (the cheap model) producing an 11.2mm f/5 full image fisheye which lacked the black corners and the flare. Framing in the finder certainly is easier when you do not have to allow for the uncertain flares. Is 11.2mm wide enough? Usually! Sharpness, unfortunately, is reduced by more than lost by cropping to the same degree and there is no avoiding the one stop loss of light. I have a comparison image that might talk you out of trying the idea. Using this extender requires taping of extender pins as shown at bottom to avoid ERR01. The extender removes the offending parts of the image (black corners and flare edges) while still giving a much wider view than the Zenitar 16mm. Taping the extender is not necessary if you use a Focus Confirmation M42/EOS adapter (eBay!) which fools the camera into thinking there is a real lens attached. Below is an example of this combination.

In summary, the Peleng 8mm Fisheye is far from perfect but can provide a fun experience and decent image quality with a little effort (and stopping down a couple stops). Certainly I would prefer a better lens with autofocus and, especially, auto diaphragm but my needs for this specialized type of lens barely justifies the price of the Peleng let alone the lens of my dreams. I recommend its purchase to anyone in a similar situation.

This article and all images are copyright © Doug Smith