Extreme video completion, where only a fraction of pixels is retained, enables energy-effective sampling. The transmitter requires no encoder hardware, close to no encoding computation, and thus no energy consumption beside the transmission. However, the reconstruction becomes challenging. Our patent covers an extreme video completion apparatus. We analyze a motion estimate with color divergence that is suitable for sparse videos. Our algorithm then exploits the estimate to adapt its spatio-temporal filter and reconstruct the video.
- Field: Computer Science
- Laboratory: IVRL
- Project leader: Majed El Helou
- Budget: 6’413.00 CHF
- Licensing Opportunity: LO 6.2082