GPU computing

Typical image processing algorithms (box filtering, windowed filtering, integral images, matrix calculus, etc) will be revisited for parallel implementations in CUDA with thread processing patterns that properly exploit the Graphics Processing Unit’s (GPU) memory and system architecture.

By the end of the course, the student will have implemented 2D image processing algorithms in CUDA, targeting (near) real-time applications. 

Required and Corequired knowledge and skills

C/C++ programming skills are strongly recommended (medium level).
Good understanding of 2D image processing algorithms. Typical examples are 2D imaging like radial distortion removal, image stitching and depth estimation in 3D applications, cf. the course INFO-H502.

The course follows a “learn by example” approach.
The exercises will prepare the student to the parallel implementation of a 2D image processing algorithm described in a scientific paper, e.g. depth estimation with stereo matching.

Contribution to the teaching profile

Massive multi-thread programming of a 2D imaging algorithm on a Graphical Processing Unit (GPU) for the multimedia engineer.

References, bibliography, and recommended reading

John Cheng, Max Grossman, Ty McKercher, "Professional CUDA C Programming", John Wiley & Sons, 2014.

Course notes

• Université virtuelle

Contacts

Office ULB-Solbosch UD5.007
Phone: 02/650 30 82
Email: gauthier.lafruit@ulb.be 

Campus

Solbosch

Method(s) of evaluation

• Project
• Oral presentation

Project

Oral presentation

The evaluation covers a report and an oral presentation of a 2D imaging algorithm on GPU, while also answering complementary questions related to the theory part of the course.

Mark calculation method (including weighting of intermediary marks)

The project report and its presentation on one side, and the theory Q&A on the other side count each for 50% of the final score.

Language(s) of evaluation

• english
• (if applicable french )