Tuesday 22 October 2019 16:00 – 17:00 A. Payatakes Seminar Room
“Neuro-inspired Compression of Images and Videos (NICE2)”
Dr. Effrosyni Doutsi Institute of Computer Science (ICS)
Abstract
In recent years the widespread use of multimedia devices has increased the volume of digital data. Compression algorithms are necessary to store and process this copious amount of data. Unfortunately, these algorithms are developing very slowly while at the same time their complexity is constantly increasing. For this reason, the signal processing community is looking for solutions that will improve the capabilities of compression algorithms in other disciplines, such as neuroscience. Understanding and decrypting the brain has gained scientific interest as it is considered one of the most sophisticated systems yet capable of managing and compressing large volumes of information. In this talk I will introduce our image compression architecture that is based on neuroscience models. We have developed tools that enable to encode the visual information by mimicking the spiking neurons without sacrificing the quality of the reconstruction. In addition, using machine learning-based image quality metrics, we show that the proposed compression system performs much better in terms of perception than the state-of-the-art.