Distinguished: AI-assisted video compression Mjellma Çitaku wins ARD/ZDF "Frauen + Medientechnologie" Award

Mjellma Çitaku, a student at Technische Universität Braunschweig, has been awarded a prize by the ARD.ZDF medienakademie for her Master’s thesis. The award recognises outstanding final theses by women in the field of audiovisual media production and distribution. Three 2023 award winners were selected from the pre-selection with ten nominated female academics. The award ceremony will take place on 26 October as part of the Medientage München.

Mjellma Çitaku studied Computational Sciences in Engineering and wrote her Master’s thesis under Prof. Tim Fingscheidt at the Institute for Communications Technology at TU Braunschweig on the topic of “Gelernte Videokompression mit tiefen neuronalen Netzen” (Learned video compression with deep neural networks). Thanks to Mjellma Çitaku, TU Braunschweig succeeded in winning a sponsorship award for the third time.

Video compression is of immense importance in times of video streaming—via mobile networks as well—because a large part of the available bandwidth of mobile networks is nowadays used for video content. Strong video compression methods therefore ensure more sustainable video data transmission with less energy consumption. But the video quality has to be right. This is where the work of Mjellma Çitaku comes in: She was able to show that a somewhat improved video quality is also possible with AI-assisted video compression.

Prof. Fingscheidt says: “This is exciting in that the development of classical video codecs took 20 years and more and now, in a rapid time, quality at a lower bit rate has become possible with disruptively new approaches. This promises improved video codecs in the near future”.

“As part of the master’s thesis, I extended a well-known video compression framework implemented in Python and combined it with various deep learning approaches. My master’s thesis showed that learned video compression methods offer advantages over classical video compression approaches and can actually be superior in some cases,” says award winner Çitaku, who graduated from TU Braunschweig in 2021. “Moreover, these methods are more flexible, efficient and royalty-free. For example, you can train them specifically for a specific application or adapt their architecture for more efficiency or performance. These features are what make machine learning extremely good for tackling complex problems.”

The public award ceremony will take place as part of Medientage München on Thursday 26 October 2023 at 1.30pm at the House of Communication. The winners of the first, second and third prizes will also be announced then. The awards include prize money with a total value of 10,000 euros. The patron of this year’s competition is Katja Wildermuth, Director-General of Bayerischer Rundfunk.

The other prize winners are Susanna Schwarzmann from Technische Universität Berlin and Sandra Severin from Technische Universität Ilmenau.

The idea of the sponsorship award

With the sponsorship award, the public broadcasters want to motivate talented women to deal with the field of audiovisual media production and distribution as part of their studies and research. In addition, talented women are to be provided with career-enhancing contacts in the public broadcasting corporations.

The ARD/ZDF Förderpreis will be awarded for the fifteenth time in 2023 to female graduates of universities in Germany, Austria and Switzerland. Under the motto “Meine Idee schreibt Zukunft” (“My idea writes the future”), final theses are sought that deal with current technical issues in the field of audiovisual and online media. The ARD/ZDF “Frauen + Medientechnologie” award is organised by the ARD.ZDF medienakademie, the training and further education institution of the public broadcasters in Germany.