Autopano Video - Stitching v2.0
The videos must have been synchronized before stitching. Please refer to the synchronization page.
Frames to use
In Autopano Video, the stitching of input videos is performed by the Autopano stitching engine on extracted frames.
You can choose between two options:
- Stitching "using the current position": only one frame per source video will be extracted and used for stitching. This is the fastest option but not the recommanded one if the cameras are not static. The frames used are the ones of the timeline current time. The best moment for stitchoing is:
- No motion of the cameras: when the cameras are not moving, you no more suffer of residual desynchronization, rolling shutter and motion blur.
- No close object: if you stitch videos coming from a camera rig, the nodal points of he cameras cannot be perfectly superimposed. Then, the stitching will compensate the parallax in a way very specific to the current surrounding 3D scene. The less you have close objects the less the stitching will be instant-specific.
- Visual information: the current time must not suffer of a lake of visual content due to blue sky, uniform color walls...
- Stitching "using the current selection": this is the recommanded option. The stitcher will use several frames to compute an optimal stitch. The number of frames to use per video can be modified in the general settings. If the motions have been previously computed, they will be used to choose moments with less motion.
- Stitch as preset: this is the recommanded option. Choose the preset corrensponding to your cameras. See the bellow Camera Presets chapter for more information.
- Stitch as Pano (template stitching): no computation will be done, the given .pano will be used to place the cameras. The panorama used as a template must contain the same number of videos as the current running project. Some templates are available on this tutorial: Template Stitching of standard GoPro Rig
- Stitch as... : This mode allows to stitch videos depending on custom settings to define:
focal (35 equiv)
The camera presets let you configure the stitching optimizer with advanced options. The presets let you define the lens type (fisheye or standard), initial focal and distorsion values, enable or disable their estimation, and choose their scopes.
The presets are defined by XML files. Their location is "<DataLocation>/Kolor/Presets/Cameras/" and "<DataLocation>/User/Presets/Cameras/" , where <DataLocation> is a folder depending on your system that you can find in the general settings window of Autopano Video. The Kolor directory contains the presets defined by Kolor. They can't be modified as they are re-written at each launch of Autopano Video. The User directory contains your personnal presets. This directory is empty at Autopano Video installation. If a preset file with the same name as a Kolor preset is added here, it will be used instead of the default Kolor one.
- The name attribute of the camera tag is the one appearing in Autopano Video presets list.
- The type attribute of the lens tag can be "fisheye" or "standard"
- The optimize attribute of the focal, disto and offset tags can be "yes" or "no": it tells if the parameter must be estimated or forced at specified value
- The scope attribute of the focal, disto and offset tags can be "pano" or "image"
- The value attribute is a real number picked from a nodal point calibration result with Autopano: http://www.kolor.com/wiki-en/action/view/Autopano_-_Panorama_Editor_-_Layers_Editor
As an example, here is the XML preset file for generic GoPro cameras rigs:
<camera name="GoPro"> <parameters> <lens type="fisheye"/> <focal value="16.56" optimize="yes" scope="pano"/> <disto optimize="yes" scope="pano" level="2"> <k1 value="0.087"/> <k2 value="0.029"/> <k3 value="-0.022"/> </disto> <offset optimize="yes" scope="image"/> </parameters> </camera>
- back-to-back rig: the overlap is not sufficient for estimating correct lens distorsion values. A preset can be used to force distorsion coefficients to previously calibrated values.