hugin Frequently Asked Questions

This document contains answers to questions that have been asked by users on the Ptx mailing list and elsewhere.

The answers refer to a fairly recent snapshot of hugin.

This is a FAQ for hugin, and contains mostly hugin specific questions. General Panorama Tools FAQ's:

Questions

General questions

Control point selection

Common problems encountered while creating a panorama

Panorama Postprocessing

Answers

General questions

Can hugin be used on the Mac?

While its shouldn't be too hard to create a Macintosh version, none of the core developers uses a Macintosh. We are waiting for Mac user with sufficient skills to make the necessary (minor) changes to hugin and pano tools.

Enblend 2.1 does not work with hugin? (windows only)

Enblend 2.1 cannot read the TIFF files created by hugin and nona.

The windows executable of enblend 2.1 was not compiled with support for compressed TIFF files, which nona produces. Until a new version of enblend is available from the authors, I provide a enblend build with complete TIFF support. Download it here

Control point selection

How do I add control points?

The control point editor is quite powerful, but its usage is probably not obvious on the first try. Here are some ways the developers use the Control Point panel:

  1. Selecting control points in 100% mode.

    This method needs some scrolling, if big images are used. You might want to try the fit to window zoom setting in that case. Switch to the Control Points tab, and use the following settings: 

    Zoom: 100%
 [X] auto fine tune
 [X] auto add
 [X] auto estimate
    • Click on a prominent feature in the left image. If the image pair already contains control points, hugin will try to select the point in the other image.
    • If its the first point in this pair, click near the same feature on the right image. The second point will be placed and fine tuned automatically.
    • If you are not happy with the placement, both points can be moved by dragging them to a better position. Press the "f" key to fine tune the point in a small area.

  2. Selecting control points in fit to window mode.

    I uses this mode if I need to set points on big images. Switch to the Control Points tab, and use the following settings: 

    Zoom: fit to window
 [X] auto fine tune
 [ ] auto add
 [X] auto estimate
    • Click on left image. The image is shown in 100% view.
    • Within the detailed view, click on a prominent feature. If the image pair already contains control points, hugin will try to select the point in the other image.
    • If its the first point in this pair, click near the same feature on the right image. The point will be placed and fine tuned automatically.
    • If you are not happy with the placement, both points can be moved by clicking at the desired position. Move the point close to the desired feature and press the "f" key to fine tune the point.
    • When the points are on the same feature, press the right mouse button, or press the "a" key to add the control point pair. It will then be shown in the list below the image.

How do I scroll both images at the same time?

Try pressing the shift key while moving the mouse. The control key or the middle mouse button can be used to scroll only the image under the mouse cursor

Common problems encountered while creating a panorama

In my panorama, the horizon is curved. How do I straighten it?

If the panorama looks nice but the horizon is curved, there are two ways to improve the image and straighten the horizon. First, try optimizing the view by selecting "Positions and View" as the optimization mode and run the optimizer afterwards. By clicking "Calculate Field of View" in the "Stitcher" tab and displaying the preview window afterwards you can check if the image has been improved.

If it is still curved, you have to add horizontal guide control points in the "Control Points" tab. What works best for me is to select the first image in the left half of the window and the image next to it in the right window, just as you do with normal control points. Then add a control point in some left part of the left image, where the horizon is. In the right image select a point on the horizon in the right part of the image. Once this point is added, change its type in the list below to "horiz. Line", as displayed in the image below (the red circles in the left/right part of the left/right image). You might want to switch off the auto-add and auto-estimate options while doing this to avoid naggy dialogs while adding this guide points (which I forgot in this screenshot).

Hugin screenshot displaying horizontal guide line between two images

Then, just cycle one image pair to the right and do the same thing. For some images where lots of buildings are you have to guess where the horizon is, but do not worry, you do not have to be that exact (and most people can guess the horizon quite well without seeing it, must be some kind of evolutional ability ;-).

Afterwards run the optimizer with the "Positions and View" setting, update the preview and check the result. For panoramas that have more than one horizontal line of images, you usually only have to set horizontal guides for one line.

When optimizing my 360 degree panorama it results in a circle instead of a hfov-360 panorama. How do I make hugin recognize it's a 360 degree panorama?

In some 360 degree panoramas you might encounter an optimization like this:

360 degree panorama invalidly optimized as circled pictures

This is not a bug, but a feature ;-) As the panorama tools can handle images in any orientation, once the optimizer has found a strong local minimum it might be unable to escape from it. The way of avoiding this is to guide the optimizer by giving rough picture estimates for the yaw values (as pointed out by Bruno Postle). To do this, count how many pictures are in your horizon line. For images with a horizontal field of view of 40 degrees this might be 10-12 pictures, sometimes more. Then divide 360 by the number, for example lets assume you have 12 pictures in one horizontal line, so 360 / 12 = 30. In the "Images" tab enter the yaw values in increasing order. Be careful to do this in the right order, i.e. if you took your pictures from left to right, just add 30 to each yaw value, starting with 0 for the first picture, then 30, 60, 90, etc. This does not have to be very exact, its just there to avoid the optimizer stepping into a local minima it might not be able to escape from. If you have multiple horizontal lines of images, you can do this for the other lines as well right now or try without doing so and if it fails, redo it later. Be sure to save your .pto file before optimizing.

Now optimize using "the Custom parameters below" and clear the pitch and roll and view optimizations. For the "yaw" box, select every picture but the first. Then optimize. Now you should land at a pretty good first estimate, which you can check in the preview tab after updating the preview. With this rough estimate, you can apply the normal "Position and View" optimization now.

If it still has some circles, you can try to set and optimize the pitch, too.

There are dark blends between two images, how do I remove them?

If there are visible black zones between two source images in the final output image but they are not visible in the preview, the most likely reason for that is feathering happening with the black background. This occurs if two images overlap only slightly and the feather width is set to more pixels than the overlap. It happens because feathering happens inside the image boundary, up to 'feather size' pixels into the image. If there is no overlapping picture within this border, the feathering takes place between the black background and the picture, which makes the black background shine through the image. To change this, either: (a) decrease the feather size to below the overlapping amount, or (b) increase the overlap of the two pictures. While (a) is always possible if there is any overlapping, (b) should be considered while taking the pictures in the first place.

Panorama Postprocessing

I want to use the GIMP to tweak the seams of my image, but the GIMP can't read a PSD format file.

Unfortunately, The GIMP can't read PSD formatted files generated by PTStitcher, and the multiple tiff output it produces is cumbersome to use. There are two possiblilites to work around this:

  1. Use our experimental stitcher, nona, to output to a multiple tiff format. Just follow the following procedure:

    • use hugin to create the panorama project, and choose "Multiple TIFF with mask" as output format. Save the project and start nona on the commandline:

      $ nona pano_project.pto multi_layer

    This will will produce a multi_layer.tif file, that contains all remapped images, cropped to their bounding box. This will save a lot of space, compared to a "traditional" PTStitcher layered output file, where all layers have the full panorama size.

    Unfortunately, The GIMP 1.2 and 1.3 can't load multilayer TIFF files. You have to download and compile an updated tiff plugin. It has been tested with GIMP 1.2 and GIMP 1.3.22.

  2. If this is to cumbersome or you need PTStitcher features not supported by nona, you can also use tif2xcf, to combine the multiple TIFF output into a multilayer XCF.

    Important: PTStitcher does not create the output files, if the output file contains an .tif extension. So, specify only the basename, without the extension.

    Unfortunately this wastes tons of memory because it stores each remapped image in a layer with the size of the final panorama.