How wide is wide: part 2. Fisheye vs. 12mm.

As a follow up to my previous field of view comparison, I decided to do another comparison of the AF-D Nikkor 16mm f/2.8 Fisheye and Sigma HSM 12-24mm f/4.5-5.6 @ 12mm. I used the Nikon D3 FX body this time around. (I really hated NX2 and uninstalled it after only a week; lost D700 native RAW support.) The original images were shot in uncompressed 14-bit RAW, manual exposure, f/8 @ 1/320th sec., ISO 200, manual WB = 5,000K, and focus set to infinity. I used Nikon Capture NX 1.3.4 and Photoshop CS3 w/ PanoTools 2.7 plugins to run the 16mm Fisheye distortion corrections. The exact same Nikon Standard Picture Controls were applied to each image.

12mm vs. 16mm animation.

Bottom line: a Capture NX de-fished 16mm Fisheye image is “wider” than the Sigma at 12mm (see animation above). The NX de-fished image is far better in terms of even lighting across the frame. The Sigma has got the worst corner light fall off I have ever seen! (But if you plan to do a lot of edge burning –this might be a good thing.) One thing the Sigma does well is create a very rectilinear projection. The NX de-fished image still has some barrel distortion on the edges. Consistent with my earlier comparison you can again see the color bias of both lenses –the Nikon is more blue and the Sigma is more yellow.

For the Nikkor 16mm example above, I used the Capture NX “Lens Adjustments–>Fisheye Lens” default dialog. NX automatically did it’s correction and then cropped the result to fit within the original frame size (4,256 x 2,832). You can also check a box to, “Include areas where there is no image data”. That yields this result (see below… if you “right click–>View Image” you can see larger images):

16mm de-fished with empty areas.

It is interesting to see how much more information there is on the fringe. Technically, there is even more image information but NX hits it’s limiter and then down-samples the results to fit into the default 4,256 x 2,832 rectangle. The problem with this method is a resultant loss of detail in the center of the image –where the “inward suction” was the greatest. If you do not plan to print the image very large, there is no problem. And if you take the above example; crop off the empty areas on the top and bottom –you get a decent faux panorama (see below):

16mm de-fished image: faux panorama.

But what if you do want to print your image very large? Of course the best (image/print quality) solution is to shoot a multiple image panorama with dedicated equipment. But for those of us who like to travel light, the dedicated pano gear is often too cumbersome and too heavy to bring along. In these cases there are other software solutions that work well with any (small and light) Fisheye prime lens to create a super wide faux panorama. I use the PanoTools (PT) “remap” plugin via Photoshop CS3 (see below):

16mm de-fished with PT remap.

PT remap does a much better job of straightening out the 16mm Fisheye images –although you do lose more image information on the top & bottom. More importantly, PT remap, explodes it’s transformations “outwards” rather than suck “inwards” like Capture NX. Practically speaking the above transformation, at full size, is 12,134 pixels by 3,862 pixels versus the “stock” 4,256 x 2,832. This is a gross up-sampling though. Yes, you do gain 3.8 times more interpolated pixels but technically you do not have anymore “resolution” than the original image had at the time of capture. BUT the point is: you DO retain that original resolution within the “sweet spot”. In contrast to Capture NX which sacrifices detail to fit it’s transformation inside of “the box”. The genius of the PT remap technique is that you can choose to down-sample the results into any sized box that you like.

In all fairness to Capture NX… upon close inspection of the PT rempped Fisheye image, you can see why Capture NX software engineers chose to limit their output where they did. The extreme edges, outside of the sweet spot, are so heavily skewed by the transformation that you get very little “comprehend-able” information. Again, all of these examples are poor substitutes for a high quality, stitched, multi-row panorama. But, in a pinch, they are acceptable AND better than no image. :mrgreen: