Zacharias Fisches

Image Sensitivity Assessment with Deep Neural Networks

Betreuer: Prof. Dr. Wolfgang Mulzer
Abschluss: Bachelor of Science (B.Sc.)
Abgabedatum: 25.09.2017

Kurzbeschreibung

Owing to its speed and simplicity, the peak signal to noise ratio (PSNR) is still one of the most widely used image quality metrics today. In runtime critical applications, such as live-video encoding, using a more sophisticated image quality metric can be prohibitive for encoder mode decisions. Bosse et al. 2017 improved on the PSNR’s mediocre performance by perceptually adapting the PSNR, thereby introducing the notion of distortion sensitivity.

In this work we will expand on the concept of distortion sensitivity and explore the possibilities of estimating distortion sensitivity with a data-driven approach.