Enfuse is based on a paper by Tom Mertens, Jan Kautz and Frank Van Reeth: "Exposure fusion". The implementation was done by Andrew Mihal (developer of Enblend) and the hugin team around Pablo d'Angelo

Some other similar applications have also been issued recently that are based on the same paper. Some seems to be even more feature loaded than Enfuse. For exemple another implementation of the enfuse algorithm is TuFuse (freeware) from Max Lyons, which features automatic DOF fusion (from a focus stack) and exposure fusion in one go as well as some additional options.

Front-end applications or plugins have also been released that make the use of Enfuse more easy and tweaking of the settings much more friendly. The droplets by Erik Krause are available for Windows users. There are GUIs available on Windows such as EnfuseGUI by Ingemar Bergmark or on MacOS X platforms: XFuse by Kevin Kratzke. The original Enfuse is a command-line program that of course can also be run on Linux.
I have tested this piece of software in many different ways already, but I felt like trying to do something challenging and quite new to me: night panography panorama.

My previous attempts to nocturnal photography were not really satisfactory as I could not show the crazy full dynamic range of such a scene. The highlight (moon and /or street lamps) would be blown out while the sky would still be too dark to my taste. Tries to correct this with HDR techniques and tone mapping was not really successful, probably because of my lack of experience.

I have decided the other night to give a try with Enfuse: the moon was full and high in the sky over my garden as it can be in the northern hemisphere on January 23, 2008....

This is a short account of my moonlight shooting experience which by the way, made me also discover the interest of the non-associative properties of Enfuse (details follow).

The output panographies:

QTVR full screen -First try 23-01-08 Manual remote shutter triggering, camera on a 5" tripod, cloudless sky and shooting location (too) close to the house.	QTVR full screen - Second try (next night) 24-01-08 Triggering from a laptop via USB link, atop a 14" pole, nice and light clouds, better shooting location.
Only 3 exposition settings: 4, 15, 30 sec (from camera auto-bracketing) EOS 5D; Tokina10-17@ 12-mm; f/4, ISO 400 Over-exposition of the highlights could not fully be recovered without retouch;-(	Up to 6 exposition levels: 1 , 2 , 4 , 8 , 15, 30 sec, EOS 5D; Tokina10-17@ 12-mm; f/ 5.6, ISO 800 This pano is one of the many possible interpretations: it give a virtual idea of the romantic atmosphere that was there IMHO;-)

While I was quite satisfied with the first panorama, from the lesson learnt on the first attempt, I had subsequently modified the conditions on the next night in order to possibly improve the composition of the final panorama and to possibly get a set of images with a wider overall dynamic range to "play with" Enfuse and abolish hurdles that were limiting creativity the night before.

Beside the an-correctible over-exposition of the bright highlights, because of wrong place for the shooting, I had missed on the first attempt to capture the magnificent reddish light that rises on the Western sky from the nearby Toulouse street lighting that itself contrasts with the blueish natural moonlight on the Eastern part of the sky. I got some help from the nice cloud pattern that reflected and diffused these two shades of glimmer this second night...

While the objective is principally to show test results, I hope that you shall share my feelings when viewing these panoramas.

An untold feature of Enfuse

I have read and I have myself written about the influence of the order that is selected to blend images for outputting a panorama, but I have yet to read about the associativity (with the algebrical meaning of that word) properties of Enfuse.

By "associativity" I mean the following:

Let's suppose for instance having a set of 7 photographs of the same scene that were exposed during 30, 15, 8, 4, 2, 1 and 1/100 sec (!) and let's name them respectively A, B, C, D, E, F and G.

Let's take first only A, D and G for the sake of simplicity. And then let's "enfuse" the following groups of images (each group with the same enfusing default settings):

(A+D), (A+G) then (D+G).

Let's now fusion these combinations of two images with the third missing image for each of the three possible cases:

that are ((A+D) + G), ((A+G) + D) and ((D+G) + A)

While the same three original images where used in every case, none of these output results are the same at all. To verify this statement, one can stack the different images as layers and see the difference mode under a graphic application such as Photoshop.

Furthermore,

(A+B+C+D+E+F+G) shall be different from (A+B+C+D+E+F+G) + (A) because (A) has been applied twice in the second instance.

In short: besides setting the options parameters that are listed in the Enfuse READ ME, it is also and in addition possible to modulate the "enfusion". One can tweak the output differently by associating the exposure in different order and combinations. This resembles me a painter who would used paint overlays to change the shade or hue or saturation its painting chef-d'oeuvre.

I have been using this method to get the panography shown above (second case only). I have not kept track of the final combination that I have applied:-(

You may yourself give a try with the reduced sample images that are provided below:

	A (30 sec)	B (15 sec)	C (8 sec)
D (4 sec)	E (2 sec)	F (1 sec)	G (1/100 sec)

Michel Thoby, March 15th, 2008

Links:
Michel Thoby's website