Brain-like Energy Efficient Image Coding
Acknowledgement: The BRIEFING project is funded by the French Republic, the Campus France and the Université Côte d'Azur under the framework of the "Make Our Planet Great Again (MOPGA)" call.
Video streaming is growing exponentially around the world. However, these services are associated with energy use and carbon emissions from devices, network infrastructure and data centers
According to International Energy Agency reports, watching a Netflix video consumes 0.24kWh of electricity/hour.
The electricity used by all data centers globally in 2018 was 198TWh, namely 1% of the global energy consumption.
There is a significant carbon footprint of the transmission, storage and processing high definition videos. Compression is the heart of this pipeline as it reduces the number of bits required to represent high quality signals. There are several compression challenges concerning the complexity and the power consumption that have yet to be considered by the signal processing community.
BRIEFING aims at addressing these challenges and develop a novel, energy efficient, neuromimetic video compression system that satisfies the human visual perception. We propose that emerging technologies like artificial intelligence (AI) and computational neuroscience can address the compression challenges.
BRIEFING will release a semantic video compression algorithm where (i) the bit allocation is automatically driven by the visual scene content using Machine Learning (ML) algorithms and (ii) the encoding process is achieved by adapting neuroscience models that approximate the coding efficiency of the brain.