360-degree panoramas featuring people are the most engaging, though these are also more complicated to produce: Moving subjects can only be captured within a single image; they do not take kindly to being stitched together across images.
For my panoramas, I’ve settled on shooting and stitching six photos around plus one each for the zenith and nadir. This way, people standing quite close to the camera can still be captured in their entirety, like so:
The 6+2 setup is quite conventional, and a number of different camera-lens pairings cater to it.
I began making panoramas in 2008 with a Nikon D300 and the Nikon 10.5mm DX fisheye lens. I have been happy with the results, though since around 2010 I often felt I was pushing the limits of what the camera could do. In this panorama, for example, I wish the camera had given me more dynamic range to play with; and reducing noise was a battle:
Shooting people requires a certain minimum shutter speed, and with my fish-eye lenses performing optimally at an f-number of around 8 to 11, low-light situations force higher ISOs, where my D300 was being seriously outclassed by the newer Nikon D3s.
Still, I wanted my next camera to muster significantly more than 12 million pixels, because panoramas are made to be zoomed into. And thus I came to own the new 36-megapixel Nikon D800, used in conjunction with the Nikon 16mm full-frame fisheye lens for the same 6+2 panorama setup. I chose the D800 over the 16-megapixel D4 despite the latter’s superlative high-ISO performance, because early reviews pointed to the D800′s flexibility — high resolution by default, with downsampling ably reducing noise in low-light situations.
With the D800 in hand, I’ve been curious to quantify image quality vs other Nikon models when shooting panoramas, so I took my D300 and D800, my friend Paul and his 16-megapixel D7000, and a D4 very kindly lent for the occasion by Rajala Camera in Stockholm, to the Östasiatiska Museum, to shoot a suitably dark, contrasty scene in constant light.
The D7000 and D300, being DX cameras, were kitted with the 10.5mm fisheye. The D800 was tested with both the 10.5mm and 16mm lenses (in DX and FX mode, respectively), while the D4 was only tested with the 16mm lens (since attaching the 10.5mm DX lens would result in too small an image to be usable). With the f-number set to 8 for both lenses, I took a series of identical shots with all cameras: 15 seconds at 100 ISO, 4 seconds at 400 ISO, 2 seconds at 800 ISO, 1 second at 1600 ISO and 1/2 second at 3200 ISO. That’s five different camera-lens pairings at five different ISOs, for a total of 25 comparison shots.
I processed the RAW files with Capture NX2 at constant settings to TIFFs, without noise reduction but removing chromatic aberration. Also, to equalize the luminosity across all images, when processing the RAW files I underdeveloped the D800 and D4 shots by one stop, the D7000 by 1/2 stop and the D300 by 2/3 stop. The resulting comparison image looks something this:
I ended up focusing on three specific areas (highlighted), as they allowed me to reach specific conclusions on sensor noise, image definition and lens quality, respectively. The following composite images are compiled directly from the TIFFs in Photoshop, and saved as 24-bit PNGs.
Looking at actual pixels near the center of the photograph, where the optics for both lenses are at their sharpest, we can conclude that:
- Noise levels for the D300 and D7000 are about the same across all ISOs, though the D7000 manages this with 33% more pixels than the D300.
- The D4 handily outperforms all other cameras in terms of noise, including the D800, as expected. I give the D4 approximately a 1.5-stop lead over the D800 (in both DX and FX mode). The D800 in turn has a 1-stop lead over the D7000 and D300.
When looking at a section of the image containing information at the very limit of the sensor’s resolution, we can see that:
- In terms of raw resolving power, the D800-16mm pairing is in a league of its own. The 16mm fisheye can obviously deliver the necessary sharpness to feed those 36 megapixels, to good effect.
- All 5 camera-lens pairings show colored text artefacts, even at 100 ISO, I assume due to the specific nature of the sensors’ color filter array. It left me wondering how it would look on a D800E, which omits the D800′s anti-aliasing filter.
- Looking at the camera-lens pairings in the three center columns — all of them delivering around the same resolution — the D4 easily outperforms its immediate neighbors as the ISOs begin to climb. Although the D800-16mm combo in the far-right column also gains noise as the ISOs climb, it continues to return the highest-definition image.
This model boat sits at the very edge of the image, where lens performance typically drops off versus the center. When stitching images into a panorama, these edge areas are for the most part masked out, though there will often be parts that make into the final composite. The ability to preserve sharpness in these edge areas is therefore important to the overal technical quality of a panorama.
What can we conclude here?
- The 16mm fisheye clearly outperforms the 10.5mm fisheye. When paired with the D300 at 100 ISO, the 10.5mm does appear almost as sharp as the D4-16mm combo, but that’s because the D300 only resolves 12 million pixels. The 10.5mm’s shortcomings are more obvious when used with the D7000 and D800, on account of these cameras’ higher pixel counts.
- The 16mm can ably sustain the D800′s 36 megapixels, even along its edges, though perhaps not with total crispness. While still providing more information than the D4-16mm pairing, I suspect the 16mm reaches the limit of its abilities with the D800. I wonder if a D800E would be able to eke out some extra detail here; my hunch is not.
D4 vs D800:
Having taken all these shots, I was curious how the D4-16mm would stack up against the D800-16mm in terms of noise and definition at 3200 ISO, if we downsample the D800′s 36-megapixel output to the same 16 megapixels as the D4.
In this composite image, The rightmost three columns contain images derived from the same original, but processed differently. Noise reduction in the rightmost two columns was conducted with Nik Software’s Dfine 2 plugin for Photoshop, but only slightly, at around 40% of maximum, to match the D4′s unprocessed output as closely as possible.
The results are mixed:
- At 3200 ISO, the D4 still shows slightly lower noise than the downsampled D800 image. Applying some noise reduction to the D800 image (before or after downsampling) brings the noise in line with that of the D4, though with a residual loss of nuance, for example in the shades of gray on the cap at the top. I suspect this is because the D4 manages to preserve more dynamic range at the ISOs begin to climb. (DxOMark tests give the D4 about a 1-stop advantage at 3200 ISO.)
- In terms of definition, however, the downsampled D800 has the advantage. The text in the lower part of the composite image is borderline-legible with the D800, but not quite with the D4. Also worth noting is that the colored text artefacts have completely disappeared from the D800′s downsampled output, whereas the D4 will need further post-processing to remove them.
- While the D300-10.5mm pairing evenly matches lens quality with the demands of the sensor, the D7000-10.5mm and D800-10.5mm pairings do not, with the 10.5mm lens proving too soft along the edges.
- The D800-16mm pairing offers the highest resolving power. Even when downsampled to match the D4′s output, the D800 maintains an edge.
- The D4-16mm pairing offers the best low-light versatility, holding on to tonal gradations longer as ISO increases.
In sum, no laws of physics were broken. If you want to check these conclusions yourself, I have the original TIFFs and NEFs available for download via my DropBox. Note, there’s 5.5 GB of files in total, though you can download individual files.
What next? I was surprised by how usable the D800′s output still is at 3200 ISO, given a little noise reduction. Perhaps in subsequent tests between the D4 and D800 it would be interesting to push the ISOs even higher, to see where the D800′s higher noise completely erodes its resolution advantage over the D4.
Then there is room for a sensor that bridges the strengths of these two cameras; the Canon 5D MkIII may be in that spot, and a rumored 24-megapixel Nikon D600 FX camera may join it there. It would be interesting to expand the scope of this comparison to other brands, and to other lens pairings.
And finally, this test doesn’t look at whether the D800 or the D800E is the better camera for panoramas, though I suspect the limitations of the 16mm fisheye may be the defining factor for that question.