Some of my colleagues at work have been asking me this question and I’ve been responding by vaguely saying the right words and waving my hands around. So I’ve done what they should have done (and I should have done before them!) and gone to the interweb for the relevant information. The first hit is the Panotools wiki page on Enfuse.
Enfuse is based on a paper written by Tom Mertens, Jan Kautz and Frank Van Reeth, in which you’ll find much more detail than I’ll be providing here. As the page explains, the basic premise to enfuse is to compare overlapping input pixels and to filter the best information from among them through to the output pixels. So how does it choose the “best” ones. It uses criteria based on exposure, saturation and contrast.
- exposure : it chooses those pixels which are closest to the middle of the range as those are considered the best exposed;
- saturation : here, enfuse favours the most highly saturated pixels;
- contrast : again, those pixels which are going to provide the highest contrast are favoured. It uses the local standard deviation as a metric for contrast.
On top of all this, the problem that the Enfuse algorithm needs to solve is to arrive at the best solution that satisfies these criteria and produces a smooth resultant image.
In the Lightroom plugin to Enfuse, the relative importance of these three criteria can be controlled with sliders in a window. Shown below are the recommended defaults that I’ve been using, and haven’t ventured from. I need to explore these a bit more (for instance, why is contrast given a zero weight?). There are also advanced options to investigate, but may or may not actually prove useful.
Expect more as and when I learn it. In the meantime, for comparison you can see the original image to that above of the Halema’uma’u plume, heading towards Hilo and as seen from Mauna Kea, in this earlier post.