What is the proper procedure for stitching photographs using Nodal Ninja and Hugin?

Question

I have been using the Nodal Ninja pano head together with Hugin for quite a while now, but I found that I get really bogged down when I make panoramas on the larger side.. let's say the sizes I am considering are about 6-shots by 4-rows: large, but not exactly gigapixel-size. Additionally, I frequently use exposure bracketing so that is also. Since I typically have about 10 panoramic shots (each with about 20-30 photographs to stitch), I would prefer not to have to manually select control points.

Nodal Ninja allows me to more-or-less reliably set certain parameters: using the RD16 rotator gives me the ability to say that every photo is rotated by 15 degrees, plus/minus small error. While this is not huge, it at least suggests initial positions for photos.

Hugin has quite a few useful things: it can process photographs arranged in stacks, for example, helping out with exposure bracketing. It also offers multiple methods for generating control points, but my experience thus far was always that those points end up being in ridiculous places: in the middle of a featureless wall, for example. Manually I would select points to lie in corners of windows or other places where two high-contrast edges intersect. Additionally, I have never been able to use the fact that I have a good guess at the initial positioning of images, because the only way of entering this data is manually though a very difficult interface.

My questions are:

1. Is there any way to easily specify the initial photograph placement? Can this be subsequently used for improving automatic control point placement?

2. Which of the many methods available in Hugin should one use to generate control points? What custom parameters should be used?

3. Should any specific protocol be followed when stitching?

Answer 1

This isn't a comprehensive answer to your question but here a few things I've found to help give good results with Hugin. I'm no guru but I've used it for around four years now and got some decent results. If you can put any or all of these into action you will need to make very few changes, if any, from Hugin's default settings.

1. If you have access to Adobe Camera Raw and have a profile available for your lens in the Lens Correction tab, use it. This will help reduce Hugin's workload by ironing out geometric distortions that are otherwise going to need compensating for, either by just letting Hugin get on with it during alignment or by using Hugin's camera and lens correction page. Most of the hard work is already done for you if you use ACR. If you're using a decent prime with low distortion this step isn't so important, but it's worth performing with zooms. This leads us onto...

2. Use a prime lens if you can as these will generally have better edge and corner performance than zooms. Control points are always in overlapping regions and are frequently near edges or in the corners. The better the lens performance in these areas the greater the accuracy of control point auto-detection. Using an aperture around the f8 mark often gives the best edge/corner quality. As for choice of lens, you don't need to break the bank as most manufacturers basic 50mm lenses offer great picture quality thanks to their simple design, and are well suited to panoramic photography.

3. While it's a little time-consuming I find it's worth manually getting rid of all control points in sky areas as they don't make good reference points (same goes for any scene with moving elements). The Celeste function can help but it's not fool-proof, and definitely not as effective as manual CP removal.

4. If you're going down the HDR route with your bracketed images I'd suggest processing your bracketed images into HDR using your preferred application, rather than using Hugin's built-in HDR functions. Your mileage will definitely vary with the built-in tools and I find using a familiar HDR application gives much more predictable results (and a much finer degree of tone-mapping control too).

Addressing your specific questions, the Yaw setting in the Image tab is the one you need if you want to specify the degree of rotation when using your pano head. Select a central image as your anchor, with yaw and pitch values of 0. Images to the left will have -yaw values and those to the right will be +yaw. If you're going multi-row, +pitch values are up and -pitch are obviously going to be down.

Regarding control points, I've never had need to look at other CP creation algorithms beyond the default cpfind as it works very well if given a decent set of images.

As for stitching protocols, if you mean Projection Type then it's a case of experimenting and going with whatever gives a good looking end result.