Seeing Things in Random-Dot Videos - Report - Illustration 6

Описание: Illustration for the technical report Seeing Things in Random-Dot Videos - 6

Degraded videos with background parameter pb greater than 0.5: we are in the "inversion domain". There are more white points than black points, therefore for the human brain white is the background and black the foreground. This is equivalent to keeping white corresponding to foreground and black to background if we perform the inversion 1-I on the image and work in the dual space (thus the name "inversion domain"). In the dual space, we are now looking for an alignment of black points (i.e. background points), or in other words a lack of white points (i.e. foreground points). This suggests to consider a left-sided statistical test in the dual space for detecting the objects.

https://arxiv.org/abs/1907.12195
arXiv:1907.12195

Research conducted by Thomas Dagès, Michael Lindenbaum, Alfred M. Bruckstein

Center for Intelligent Systems
Computer Science Department
Technion, Israel Institute of Technology