Simon Deniffel

Since January 2024 I am researching learning based image coding. The main goal hereby is to compress images with the help of neural networks to reduce data rate and simultaneously keep a high visual quality. I am currently working on entropy models, which are used to estimate the statistical properties of images that have been compressed by a neural network in order to further enhance compression efficiency.

I offer theses in the field of learning based image coding, especially in the following topics:

• Classification of images for coding tool selection
• Classification of images for generating specialized training datasets
• Efficient training using very small datasets
• Optimization of entropy model architectures