This thesis can be done in our office in Gothenburg, Stockholm or Sundsvall
High level description
Though images provide numerous benefits to humanity, they also produce issues of their own. If one would sum the number of resources spent on transferring, compressing, or decompressing images. One would not only be surprised at the amount of energy in terms of wattage used to display images or train machine learning models; but shocked at the amount of power that is allocated to decompress, compress or send image data through the wire.
Project Description
Explore several popular images or video formats and what can be done to reduce the size or compression/decompression performance of the image. There are plenty of unexplored options in plenty of popular compression algorithms or even unnecessary data that is stored in the image (which can be removed or improved).
There's also the possibility of exploring the usage of AI to generate models trained for compressing images and how they perform compared to current compression algorithms.
Who are we looking for?
Bachelor in Software Engineering or Master in Computer Science/Engineering related field.
Purpose and scope
Bachelor level:
- Either create or approve existing compression algorithms.
- Explore what chunks of data can be removed from existing image formats.
- Measure performance metrics in said cases and compare them to the prior art.
Master level:
- Explore state-of-the-art machine learning in relevant fields.
- Generate a machine learning model that compresses images using state of the art.
- Measure performance and summarize the state-of-the-art in relevant fields.