This presentation proposes ideas and projects for improving quality, reducing cost and maximising flexibility by processing one or more video streams after recording. Ideas include:
1. Using several fixed cameras as an alternative to more expensive tracking PTZ cameras, using virtual panning or mixing on multiple recorded videos to produce a single video focused on the lecturer or region of interest
2. Adjusting for more consistent brightness, contrast and colour
Matterhorn 2014 Unconference: Ideas for automated post-recording video handling
1. Matterhorn Unconference 2014
Let’s fix it in post! Ideas for automated
post-recording video handling
Stephen Marquard
University of Cape Town
19 March 2014
2. Why post-process video?
A different approach to solving the blackboard challenge
Improve video quality
Close-ups
More natural perspectives
Avoid challenging lighting conditions (e.g. projection)
Smooth motion (perfect knowledge of the future)
Adjust for optimal colour / white balance / contrast
Cost
Multiple fixed HD IP cameras for half the cost of a PTZ camera / solution
6. A real-time blackboard solution
LectureSight:
Direct a PTZ camera to
follow the presenter, using
real-time image analysis
from an overview camera.
https://opencast.jira.com/wi
ki/display/LECTURESIGHT
/LectureSight+Project+Ho
me
7. A post-processing presenter solution
http://wiki.teltek.es/display/GaliTracker/GaliTracker+Home
GaliTracker
does
cropping and
zooming
from a fixed
HD image to
give a
presenter
closeup
11. Post-processing
The capture agent records multiple camera streams
(probably IP cameras): the mediapackage has 2, 3 or 4
presenter videos rather than 1.
The camera streams are then post-processed to create a
single presenter video (perhaps on the CA itself)
Approaches:
1. Virtual panning (combine and crop)
2. Mixing (source selection)
12. Virtual panning
Left camera
frame
Centre camera
frame
Right camera
frame
Single panoramic image constructed from 3 fixed camera images
Wide frame is cropped to position the presenter optimally (two-pass approach)
Pros: smooth motion, optimal view.
Cons: stitching images may be tricky, shifting perspectives.
13. Mixing from multiple sources
Image analysis and two-pass approach to select transition points between
cameras (switch or cross-fade).
Left camera Centre camera Right camera
Selected
camera
14. Other post-processing ideas
Improve colour, white balance and contrast dynamically or
per region in the image.
Compensate for the
effect of projector
image in the frame,
and poor presenter
lighting.
15. Other post-processing ideas
Improve image perspective and geometry
Correct for fishbowl
and other disturbing
lens and camera
positioning artefacts
16. What next?
• Get involved! Good project topics for students.
• Share your ideas and possible contributions on the
Matterhorn list(s).