Meaning of Verticle/Horizontal control point distances

[Judson Fisher]

Hi,

I'm enjoying very much getting to learn about how Hugin works, one thing that has not been clear to me is:  What is the significance of control point distance in regards to horizontal and vertical control points?

Thanks,

/Jud

[Abrimaal]

The vertical and horizontal lines are used for calculating a good looking panorama. The distance is important to avoid errors in calculations.

The V lines should be distant as far as possible, at least two near the edges of a panorama, to calculate and straighten the view.
When there are lines only at one side, the other side of the panorama may be curved up or down (for example when there is only one straight object near the left or right edge, the rest is landscape).
It is better to add a horizontal line to compensate the curved horizon (or a vertical line on an object that should be straight, for example a tall tree).
When the panorama is a landscape only, the detector may place vertical lines on trees, grass and other objects. If it looks good, leave as it is, but try also the [Straighten] button to compare how it looks without the detected lines.

Why the distance is important: to avoid statistic errors. The detector may place 5 vertical lines on the same traffic sign or electric pole that is not straight (and repeat the error on objects arranged at similar angle).
The best to place vertical lines only on firm straight objects, such as buildings, windows in buildings etc. The detector finds high contrasts, but it does not recognize if this is a building, a tree or a human leg.

What are horizontal lines for: when you work (for example) with front views of architecture, they are used to optimize the left-right position and barrel distortion. They are not detected, not introduced yet. Finding them is the same work for the detector, it is equally simple. It's hard to say why the developers did not introduce this yet.

The distance between the horizontal lines is the answer for "what you want to straighten". If you want to straighten the whole image, they should be as far as possible. If you want to straighten a specific object, a front view of a building, a photographed picture, a single window, the lines should be placed only on this object, the rest is not important.
"a photographed picture" refers not only to a single photo of a picture on a wall or a banner on a street that you want to use,

it refers also to panoramas of maps and other large sheets, scanned or photographed.

The horizontal lines are also useful and the only way to straighten the horizon when you work with landscapes or waterscapes (and there are no human-built objects).

[Judson Fisher]

Thanks for the informative description of the use of line type control points. Unfortunately my question was not specific enough, as this did not answer my question which is:

What is the meaning of the value in the "Distance" column of the Control Point Table for entries that are horizontal or vertical lines?

For instance, in normal control points, the distance refers to the distance between two corresponding points in (presumably) different images, where the goal is to reduce this distance so that the corresponding points are perfectly aligned.

But in horiz. and vert. type control points, you select not corresponding points on separate images, but two different points on the same image.  It would seem then that the meaning of the value found in the Distance column of the Control Point Table for these kinds of control points cannot be interpreted in the same way as for normal control points.

Furthermore, I'm am curious if the horiz. and vert. control point distances factor into the calculation of mean control point error as displayed in the prompt to accept the results of an optimization.

thanks for your help,

/Jud

[T. Modes]

Am Freitag, 10. Januar 2020 01:01:43 UTC+1 schrieb Judson Fisher:

What is the meaning of the value in the "Distance" column of the Control Point Table for entries that are horizontal or vertical lines?

Both points are transformed to the output space. The distance is then the (maximum) distance of the line between both points and the horizontal/vertical (in the output projection).

 

Furthermore, I'm am curious if the horiz. and vert. control point distances factor into the calculation of mean control point error as displayed in the prompt to accept the results of an optimization.

They are included in the mean control point error. (There is no differentiation between normal and line control points in the mean calculation.)

Thomas

[Judson Fisher]

Of course, I see, the maximum distance. That makes sense. Thanks for the clarification.

/Jud

--
A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to a topic in the Google Groups "hugin and other free panoramic software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hugin-ptx/x7yQ7nxUh_8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hugin-ptx+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/184cd8f3-6c0c-4542-937f-5b38f9657e96%40googlegroups.com.