The purpose of  the Image Analysis and Computer Vision course is to study both the foundations on image formation, image analysis, 3D scene reconstruction, and the methodologies underlying the solution techinques to the main problems involved. 

Image analysis addresses the extraction of the content of one or several images; Computer Vision addresses the methodologies to obtain a representaton of the observed 3D scene. Optical aspects, geometrical ones, and algorithmic aspects are studied, as well as aspects connected to signal processing and data analysis. 

A project allows to closely examine, possibly from a practical point of view, one or more of the discussed topics.

1. Knowledge and understanding: Geometric foundations of the image projection; Image analysis as a signal

2. Applying knowledge and understanding: Use geometric knowledge to devise techniques to interpret an observed scene; Use knowledge on signal processing to develop techniques to extract relevant features from images

3. Making judgements: Given a complex and realistic scene reconstruction problem, try to infer a sufficient set of constraints from the particular circumstances of the problem instance

4. Communication skills: presenting own ideas to the teacher/gruop mates relative to a complex problem to solve, reporting adopted solution and results

5. Learning skills: learning to correlate items from an eterogeneous corpus of knowledge to solve a complex and realistic Computer Vision problem, use foundations to allow lifetime learning in Image Analysis and Computer Vision

0. Introduction.

1. Camera sensors: transduction, optics, geometry, distortion

2. Basics on Projective geometry: modelling basic primitives (points, lines, planes, conic sections, quadric surfaces) and projective spatial transformations and projections.

3. Camera geometry, and single view analysis: calibration, image rectification, localization of 3D models.

4. Multi-view analysis: 3D shape reconstruction, self-calibration, 3D scene understanding.

5. Linear filters and convolutions, space-invariant filters, Fourier Transform, sampling and aliasing. 

6. Nonlinear filters: image morphology and morphology operators (dilate, erode, open, close), median filters.

7. Edge detection and feature detection techniques. feature matching and feature tracking along image sequences.

8. Image segmentation, contour segmentation, clustering, Hough Transform, Ransac (random sample consensus). 

9. Object tracking, object recognition, classification and learning.

Basics in: Physics, Calculus, Algebra, Geometry, Programming, and Algorithms

Evaluation. The examination consists of an ORAL PROOF, which is constituted by two parts: theory and application. A written, individual homework can (possibly partially) substitute the "theory" part of the oral proof. The "application" part of the oral proof consists in the discussion of a project on a topic from a list proposed by the teacher (project can be extended to Master Thesis).

A. Written individual homework: study the geometric aspect of the assigned problem, exploit problem specification to derive useful constraints  (1,2, 3); implement the solution (including the extraction of the needed image features), while addressing noise and uncertainty on the extracted image features, and evaluate experimental results (3, 5); carefully write a self-sufficient report (4).

The evaluation of the  homework has three possible outcomes:

1) from good to excellent -> the theory part of the oral proof is skipped, and the  oral proof will only  concern the  project;

2) from slightly below to slightly above the decent level  -> the theory part of the oral proof will only concern the weak points of the homework

3) from undelivered to bad -> the theory part of the oral proof will concern any theoretical topic of the course (complete oral)

B. Assessment of project artifact with possible written report: Assessment of the design and experimental work done either individually or in small groups (up to three participants): 2, 3, 4, 5.

Project can be either full or small. Full projects evaluation outcome ranges up to maximum 30/30 (cum laude) in the final mark, while small project evaluation outcome ranges up to maximum 24/30 in the final mark.