Foto 7

Modifica Data del Corso: M. Corsini, Visual Computing Laboratory – ISTI – CNR, "Multi-modal Registration of Visual Data", 6, 7, 9, 10, 15 June 2016

Hours:
15 hours (4 credits)

Room:
Aula Riunioni del Dipartimento di Ingegneria dell'Informazione, via G. Caruso 16, Pisa – Ground Floor

Short Abstract:
Registering visual data of different type is becoming an important task in many applications such as appearance acquisition of real objects, video navigation, video editing, and data visualization. In this course, we focus mainly on the algorithms to align images on 3D objects/scenes. After a brief introduction about Computer Vision and Computer Graphics, some of the most used features detectors and descriptors for images and geometric data are presented. Then, the state of the art in image-geometry registration is discussed in depth and some of the most representative methods analyzed. Finally, we discuss recent studies about specific applications such as the registration of paintings/drawing on a 3D scene, the navigation of video through the alignment of their respective underlying geometric content and the trend in scene understanding by exploiting the relationships between natural images and large collections of 3D models.

Course Contents in brief:

  • Introduction
    • Goal of the course
    • Computer Graphics
      • 3D scene representation
      • Rendering
    • Computer Vision Topics
  • Features Detectors and Descriptors for Images
    • Corners and Edges
    • Scale-Invariant Features Transform (SIFT)
    • Speed-up Robust Features (SURF)
    • Histogram of Oriented Gradients (HOG)
    • Binary descriptors (BRIEF and BRISK)
  • Features Detectors and Descriptors for Geometric Data
    • Spin-images
    • MeshDOG e MeshHOG
    • SIFT-inspired (meshSIFT e LD-SIFT)
    • Heat Kernel Signature (HKS)
    • Range maps alignment
  • Image-geometry registration
    • Problem formulation
    • Fixed-relative methods
    • Feature-based methods
    • Statistical methods
    • Multi-View methods
  • Recent advances and applications
    • Painting-to-3D scene registration
    • Video-to-geometry registration
    • Joint properties of images and 3D models for scene understanding

Schedule:

  • Day 1 (10:00 - 13:00) – Introduction
  • Day 2 (10:00 - 13:00) – Features based detectors and descriptors for images
  • Day 3 (10:00 - 13:00) – Features based detectors and descriptors for geometric data
  • Day 4 (10:00 - 13:00) – Image-geometry registration
  • Day 5 (10:00 - 13:00) – Recent advance and applications