Graph-based transforms for image processing

This activity deals with the field of graph signal processing. In particular, we design novel transforms for image processing based on the framework of the Graph Fourier Transform. We have developed a new directional transform, called Steerable DCT, which extends the discrete cosine transform by rotating ita 2D basis functions; this framework can indeed be applied to any linear transform. Similarly, we have proposed a Steerable Discrete Fourier Transform. We have applied these transforms to image/video compression, including integration in the HEVC test model.

ERC Sector:

  • PE7_7 Signal processing


  • Data compression
  • Transform coding

Research groups