AI-based representations of visual media


Recent advances in the field of AI have provided extremely efficient tools for media representation and processing, including generative adversarial networks, implicit neural representations, neural radiance fields, diffusion models, graph convolution, and so on. These representations are extremely flexible and can be adapted to heterogeneous visual media like images, 3D scenes, point clouds, and other types of multidimensional data, enabling the solution of various problems that were extremely complex to tackle with traditional signal processing tools. Research directions include the solution to inverse problems in irregular domains such as denoising, super-resolution, and restoration of point clouds, the efficient rendering of multidimensional scenes, joint-representations of different media like images and point clouds, just to mention some.

ERC sectors 

  • PE7_7 Signal processing
  • PE7_8 Networks, e.g. communication networks and nodes, Internet of Things, sensor networks, networks of robots
  • PE6_2 Distributed systems, parallel computing, sensor networks, cyber-physical systems
  • PE6_5 Security, privacy, cryptology, quantum cryptography
  • PE6_7 Artificial intelligence, intelligent systems, natural language processing
  • PE6_8 Computer graphics, computer vision, multimedia, computer games
  • PE6_11 Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)


  • Generative adversarial networks
  • Neural radiance fields
  • Diffusion models
  • Graph signal processing
  • Point clouds