Subject

Recent work in computer graphics has demonstrated impressive image quality for generating photo-realistic images from arbitrary point of views. The method is based on photogrammetry, where a small number of photograms are captured at different point of views and any view can be interpolated from the plenoptic data representation.

The immersive plenoptic image representation is capturing all possible light rays we could see. based on photographic pictures instead of classical modeling of the geometry and textures. This allows for capturing complex light-matter interactions such as frosted glass and complex reflections.

Kind of work

We are interested in exploring further plenoptic imaging with different kind of Master Thesis topics, according to the interest of the students. Some possible topics could focus on the following propositions:

1. Technical topics could implement the technique, starting from existing open-source examples and several improvements may be experimented, under the guidance of the supervisor.
2. We are interested in evaluating the perceived image quality with both automatic quality evaluation metrics, as well as subjective methods. The freedom of viewpoint selection poses new challenges for grasping an overall perceived quality.
3. We are interested in implementing a Gaussian Splatting method on the GPU, for real-time rendering of views. There are possible improvements for modeling complex light transmission and reflection effects, adding to realism.

Framework of the Thesis

You will integrate the group of Peter Schelkens at the ETRO department of the VUB where facilities are available, such as our Visualization Lab and computers with advanced GPUs. Your promotors will be Dr. Saeed Mahmoudpour and Pr. Colas Schretter depending on the topic of interest. We are located in Room 26, second floor of the Pleinlaan 9 Building and you are welcome to meet us for discussing more in details the topics.

Number of Students

1