Image and volume data analysis (2018/2019)

Course code
Andrea Giachetti
Academic sector
Language of instruction
Teaching is organised as follows:
Activity Credits Period Academic staff Timetable
Teoria 5 I semestre Andrea Giachetti

Go to lesson schedule

Laboratorio 1 I semestre Andrea Giachetti

Go to lesson schedule

Learning outcomes

The course aims to provide the knowledge necessary to understand and use algorithms to process digital images and different types of spatially related data (volumes and surfaces). Therefore, algorithms and data structures will be presented to effectively code the data, segment the regions of interest, characterize with descriptors, recognize objects and align the structures (registration).

At the end of the course, the student will have to demonstrate knowledge and understanding skills that allow him to exploit data acquired by multimodal probes to perform 3D reconstruction, measurement, recognition and information fusion.

In addition, the student must demonstrate that he is able to use notions of computational geometry, algebra, and algorithms on graphs to solve practical problems in various application contexts, autonomously select the most appropriate data structures and the best algorithms.

The student will then be able to present an application project describing effectively motivations and choices, continuing the studies independently in the Visual Computing domain.


1 Introduction
Spatially referenced data in 2D and 3D, scattered data, gridded data, sensors
2 Data structures
Images and volumes, binary and label data, topology. Scattered data, triangulation and interpolation
Surface and contour representations: curve approximations, splines; point clouds, meshes, meshing algorithms Moving Least Squares, ball pivoting.
3 Segmentation
Classification-based approaches, graph based approaches,
4 Contour/surface based segmentation
Snakes and active surfaces 3D, Level Sets, fast marching algorithm, distance maps and geodesic distances
5 Model based segmentation ,
Model fitting Hough transform, statistical modelling with training data: Point Distribution Models
6 Shape analysis
Moments, binary region descriptors, invariance properties.
Contour based descriptors, local features, signatures, Fourier descriptors, context in local descriptors.
Skeletons and medial lines,
Differential geometry of surfaces, descriptors on meshes
7 Image and texture features
8 Spatial registration
Problem descriptions and approaches. Landmarks based registration,
Intensity based registration

Laboratorio Matlab
Introduzione a Matlab, images, contours, sampling and interpolation, splines
Voxel based segmentation
2D Snake implementation
2D Shape retrieval
Mesh processing, 3D shape retrieval

Assessment methods and criteria

Written test (20/30) and evaluation of programming skills (10/30)

To pass the exam, the student must show
- they have understood the principles related to surface and volume Data encoding and processing
- they are able to describe these concepts in a clear and exhaustive way
- they are able to apply the acquired knowledge in different applicative contexts

Written test:
The written test is composed by a few open questions and/or exercises testing the understanding of the different topics of the course.

The lab exam will consist of a small project applying methods learned in the course to application specific data