Jump to: navigation, search

Optimizer engine

Unknown sharing


The role of the optimizer is to estimate all the unknowns of the panorama (except color unknowns that are treated separately). There are two types of unknowns:

  • The position of each image: yaw , pitch, roll
  • The calibration of devices used: focal length, distortion coefficients ( k1 , k2, k3 ) and the centering of the optical axis (offset X and Y)

Most of the time, several images of a panorama (usually all) are from the same camera, set the same way. In this case, the calibration unknowns are shared by multiple images. The concept of scope allows this sharing. Sharing unknowns has two important advantages:

  • Reduce number of values to estimate and therefore the work performed by the optimizer.
  • Use more information to better estimate each value. Indeed, if a calibration unknown is shared by all the images in a panorama, then every match between any images participate in its estimation. This is very important because in most cases, a single pair of images is not enough to properly calibrate a device.

It is therefore important to share the unknown when possible.

In practice

Setting scope

In practice, in Autopano, each unknown has a scope which can be:

  • Image: the unknown to be estimated is specific to each image. Typically, it is the case of positions (yaw , pitch, roll)
  • Group of images: the unknown is common to multiple images but not all images of the panorama.
  • Panorama: the unknown is shared by all the images of the panorama. This is most often the case for the unknown calibration (k1 , k2, k3, offset X , offset Y)

Creating groups of images

There are three ways to create groups of images:

  • In automatic mode, Autopano itself creates groups based on EXIF information. In the absence of EXIF information, Autopano can also make the decision to group images if their estimated calibrations are very close.
  • Give the creteria for grouping among the five possible: type of lens, stacks, camera model, focal length, presence of EXIF
  • Manually edit them into the "Optimization groups" window of control points editor

Examples of groups and scopes configuration

Grouping criteria Scope of yaw, pitch, roll Scope of focal Scope of distorsion Scope of offset X & Y
Multiple focal length stitching focal length image group group group
Livepano presence of EXIF image group group group
360° video rig No grouping image pano pano image
HDR stack group pano pano pano

Result example

Without unknown sharing

Optimization settings



ResScopes1 80.gif

With unknown sharing

Optimization settings



ResScopes2 80.gif

Source Images by Arnaud Frich

BACK TO: Documentation / Autopano Documentation