DPR Forum

Welcome to the Friendly Aisles!
DPRF is a spin-off of dpreview. We are a photography forum with people from all over the world freely sharing their knowledge and love of photography. Everybody is welcome, from beginners to the experienced professional. From smartphone to Medium Format.

DPRF is a community for everybody, every brand and every sensor format. Digital and film.
Enjoy this modern, easy to use software. Look also at our Reviews & Gallery!

Mosaics: Pros and Cons

AlenK

Active Member
In the spirit of sparking new discussions, let's talk mosaics, or as most people seem to know them, panoramas. Personally, I prefer "mosaic" to "panorama" because the latter was historically applied to image frames that were significantly wider than they are tall whereas the former covers image frames of any aspect ratio. (It is not for nothing that several widescreen film formats first introduced in the 1950's and the 1960's have "rama" in their names.)

For the uninitiated, a mosaic is a series of overlapping frames, typically arranged in a rectangular grid of MxN panels. For a traditional horizontal panorama, N (the vertical dimension) would be 1. Some overlap is necessary between adjacent panels to ensure that they can be properly stitched together with software.

I am a big fan of mosaics. Why? Primarily for two reasons. One, because I bought an APS-C DSLR and two, because I am naturally thrifty. Back in the day, I had been using a 35mm film SLR. Moving to an APS-C DSLR meant that I was no longer getting nearly as wide a field of view (FOV) through my telescope nor through the lenses I already owned. That was good for some deep-sky objects (the smaller ones) and not so good for others (the bigger ones). For wide-field images through lenses, it meant that an 18mm focal length, for example, was only giving me a FOV roughly equal to the FOV from a 28mm lens on a full-frame camera (which of course is what a 35mm film camera is).

This is where the thrifty part comes in. Doing a mosaic is much less expensive than buying new optics, whether telescopes or lenses, that would reclaim the wider fields of view I had with a full-frame camera and less expensive than buying a full-frame digital camera. In fact, when it comes to lenses, I have found that mosaics using an inexpensive lens give just as good an image as one taken with a more expensive lens with an equivalent FOV. That is due to image scale and aperture.

Because a mosaic uses multiple exposures, it can obviously reproduce a given FOV with higher pixel resolution than a wider-angle lens rendering that same FOV on the same camera without using a mosaic. That also means that the aberrations the lens creates when it focuses a star are rendered smaller, making them less obvious than in a single image taken with the same lens.

Even if the mosaic image is downsized to, say, match the resolution of the non-mosaic image taken with the wider-angle lens (an apples-to-apples comparison), the improvement in the rendering of stars is largely maintained. It’s as if a better (usually more expensive) lens has been used than was actually the case.

Aperture is the second reason mosaics taken with an inexpensive lens can match a non-mosaic image taken with a better, wider-angle lens. At the same f-ratio, the longer focal length lens has a larger aperture diameter. It is aperture diameter, not f-ratio per se, that determines how much light a lens can collect. And it is primarily how much light you collect that determines the quality of an astrophoto (provided everything else is in order, such as focus). Even if a slower (higher f-ratio) lens is used to do the mosaic, it could still have a larger aperture diameter than a shorter focal length lens.

So what are the downsides to a mosaic compared to a single frame image? For one, it is more difficult to take since you must re-aim the camera for each panel in such a way that you get the desired degree of overlap. This gets easier with practice and there is hardware that can help, such as an indexed panning head for your tripod.

The second downside is increased time and effort at the image processing stage. You must now process MxN frames, or MxN stacks, rather than just one frame or one stack of frames. There is really no way around this.

But I would argue that the upsides of mosaics more than justify the downsides. Supporting this view is the high quality of the results obtained by using mosaics that can be found on the internet. There are now many people posting impressive mosaics. My favorites are from Eric Benedetti out of Utah. He posts mostly nightscapes (Utah is blessed with an abundance of interesting foregrounds) but also does deep-sky images with lenses. He does mosaics almost exclusively now.

So, have you tried taking mosaics? If you have, how did they go?
 
I've done terrestrial panoramas but I have yet to shoot my first night sky. For the terrestrials I've done my own matching which has usually involved some color adjustments. I wouldn't expect to have to do that with night sky shots.
 
I love to do terrestrial mosaics too. I have posted some of my best here (a link that will eventually break when DPReview finally disappears). I haven’t had to make any color adjustments to panels but I have experienced some stitching artifacts from time to time. I could have fixed those with more effort.

All of my terrestrial mosaics were hand held, which may have contributing to the stitching difficulties. Use of a nodal tripod head is the only sure way to avoid them. Fortunately, one of those is usually not necessary for astronomical mosaics since everything in the night sky is at infinity (as far as the camera is concerned) and terrestrial foregrounds are typically far away from the camera.

My very first astro-mosaics were taken on the same night:

ROXEdBgE0-_s_16536x16536_JXczOqwr.jpg

(Details: https://www.astrobin.com/1b4518/0/)

HTvei4KNvx8B_16536x16536_kWXURFLk.jpg

(Details: https://www.astrobin.com/7gm3f6/0/)

The second, irregular mosaic/panorama was accidental. I was just testing the Astrotracer internal tracking feature of my Pentax K-3II and realized when I got home that I had several overlapping frames centered around that very-interesting area of the Milky Way.

Although neither image is what I would call impressive, the fact that I managed my first and second astro-mosaics on what was my first serious night of using that camera speaks to how easy they can be to do. I should qualify that as “how easy they can be to do with Astrotracer.” They are just as easy to do with an untracked camera on a tripod but not quite as easy to do with a tracking mount because in the latter case, movement along the RA and Dec axes complicates alignment of the panels.
 
Continuing on in this thread, I wanted to post an analysis of the 3x3 nightscape mosaic I posted above. I used a vintage SMC Pentax-M 28mm f/2.8 lens that I have had since my film days.

Each exposure was 60 seconds with the lens wide open. (I normally close the aperture down one stop to f/4 but not on this occasion.) The exposures we’re tracked using the camera’s internal Astrotracer feature. The overlap between adjacent panels was roughly 50%. “Roughly” because I aimed the shots by eye through the viewfinder, not by using a calculated rotation angle on an an indexed tripod head.

Given the amount of overlap, the effective total exposure time does not likely exceed 2.5 minutes, despite having done 9 minutes of actual exposure. This is due to the excessively large overlap. The light collected in the overlap region should only be counted once because the stitching software blends (crossfades between) the panels during stitching.

There is little reason to exceed around 30% overlap and with experience closer to 20% is achievable even with wide-angle lenses. The less overlap, the less light loss due to blending and, of course, the wider the final image.

I believe my cropped, final image is roughly equivalent in field of view to that from a 14mm lens. Given that I used an aperture diameter of 10mm, a 14mm f/2.8 lens operated wide open (hence an aperture diameter of 5mm) could have given a similar image, including with respect to the signal-to-noise, using 10 minutes of total exposure time.
 
Last edited:
Oops, I messed up the math a bit in the prior post. The effective exposure time of the mosaic because of the overlap is actually 4 minutes, not 2.5 minutes. That means a 14mm f/2.8 lens, with 1/4 the light collection capability of a 28mm f/2.8 lens (5^2 / 10^2) would have to be exposed for a total of 16 minutes to get the same signal-to-noise ratio. That’s roughly about how long I probably took in total to take the mosaic factoring in the time it took to re-aim the camera for each successive panel (I had no experience with mosaics). So the comparison is even closer.

A used copy of that 28mm f/2.8 lens in usable mechanical and optical condition could be bought today for well less than $100US - I have seen them for close to $50US - whereas a used copy of a 14mm f/2.8 lens, say a Samyang/Rokinon/Bower model, seems to typically sell for between $150US and $250US.

So, you aren’t saving a lot of money by doing a mosaic in this case and you could argue that the ease of taking the single-frame image with a 14mm lens easily justifies its modest extra cost compared to the vintage 28mm.

Can we do better? Let’s compare instead with a 14mm f/1.8 lens. Used copies of such lenses seem to start at around $1,000US. (For Sigma Art lenses - Sony FE is around $400 more.) Such a lens would only have to be exposed for about 6.5 minutes to get the same SNR as the mosaic image. In this case, using the premium lens indeed saved some time but at a huge cost penalty.

Would the premium lens have yielded better star shapes than those in the mosaic when comparing the images under the same conditions (same image size at the same viewing distance)? Considering that the stars in the mosaic have been shrunk down considerably - probably to no larger than 30% of their linear size compared to the original image width - it’s really hard to say. A side-by-side comparison of images would be needed. My guess is that the premium lens might indeed be better but not by a lot.
 
Back
Top