Using Gigapan module
GigaPan is collaboration between Carnegie-Mellon University and NASA, with support from Google.
Its purpose is to facilitate the creation of Gigapixel panoramas, and to host them on a dedicated web site: Gigapan.org.
GiganPan Systems distributes motorized heads which automatically take photos in a grid configuration. Photo stitching software is also offered.
Autopano’s Gigapan module makes it possible to stitch photos taken by GigaPan systems.
The purpose of this tutorial is to compare Autopano’s data processing possibilities for GigaPan with existing solutions.
Photo shoots will not be discussed here (see the tutorials on the GigaPan site).
Example 1: San Francisco Bay - 1,1 Gigapixel
- Number of images: 184 photos.
- Configuration: 8 rows x 23 columns.
- Direction of shoot: by columns, from left to right.
Gigapan Import Module
In this Import module, you can choose from the various shoot layout options available with GigaPan heads and configure the steps to take afterwards (control point detection, stack management if necessary).
The default parameters are set for automatic control point detection and automatic stitching.
Stitching with Autopano Giga
This is the result of the stitching in Autopano Giga, using the default settings:
Stitching with GigaPan Stitch
There are some stitching errors on the bridge:
The same area, from the image generated with Autopano
Example 2: Print Shop - 3 Gigapixels
- Number of images: 700 photos
- Configuration: 20 rows x 35 columns
- Direction of shoot: by columns
Checking the 360° Panorama option in the Shoot Setup tab in the GigaPan import module, forces the software to link the images in the first column to the ones in the last column.
The panorama is stitched correctly, with no trouble joining the ends.
Example 3: Football Game - 2,45 Gigapixels
- Photographer: Stoney Vintson
- Hardware: Nikon D700 (12,1 Megapixels - 4256x2832)
- Number of Images: 756
- Configuration: 21 rows x 36 columns
- Direction of shoot: by columns, from bottom to top
- Link: http://gigapan.org/gigapans/35989/
This set of images shows how Anti-Ghost performs when used on Gigapixel images. We can observe differences in the quality of the stitching, which is markedly better in Autopano.
Managing images of moving people with Anti-Ghost also gives better results.
There are still some problems that Anti-Ghost can’t solve, because a person can appear partially in an area of overlap (which may or may not be retained) and partially in a non-overlapping area (which must be retained).
In some cases, depending on the type of scene to be captured, it can be wise to choose one type of acquisition configuration over another (by column or by line).
Example 4: Dubai - 45 Gigapixels
- Gerald Donovan re-rendered this panorama with the new version of Autopano Giga:
Comparison of rendering times
- Test Computer Specifications:
- 12 GB RAM
- 24 CPUs
- Data Set Used:
- 700 photos 3264 x 2448 (8 Mpixels)
- Panorama rendering at 50% (42048 x 18176 pixels)
|GigaPan Stitch 1.0.0805||Autopano Giga 2.5.1|
|Preview Display Time||
In Autopano, the user can control the stitching quality, edit the panorama if necessary, and add control points manually or in automatic mode, then start the rendering, which makes the processing very flexible.
In GigaPan, the user has no control, and the flaws in the stitching can’t be corrected.
The panorama can be edited in Autopano after the preview is displayed, which happens very fast. The user very quickly gets an idea of what the result will be.
Moreover, editing allows the user to adjust the panorama’s orientation (yaw, pitch, roll), to straighten vertical and horizontal lines, to define the extent of the rendered area, the type of projection, exposure equalization, etc.
It is also possible to export files in various standard formats (including .psd/.psb).
Rendering time is longer for Autopano, but the better results with Anti-Ghost make up for this.
The Anti-ghost algorithm was specially developed to applicable to Gigapixel images.