Corso di laurea magistrale in Informatica

Upon completion of the course, the student is able to select and parameterize algorithms suitable for improving image quality in the spatial and frequency domains, lossy and lossless image compression, and solving computer vision problems such as classification, segmentation, and image generation.

KNOWLEDGE AND UNDERSTANDING SKILLS. Acquisition of methodologies of selection, and parameterization of natural and medical image processing agorithms.

ABILITY TO APPLY KNOWLEDGE AND UNDERSTANDING. Acquisition of theoretical for natural and medical image processing.

SELF-JUDGMENT. Acquisition of the basic criteria for determining how to set up proper algorithm design for natural and medical image analysis.

COMMUNICATION SKILLS. Acquisition of technical terminolgy in the area of natural and medical image eleprocessing.

LEARNING SKILLS. Development of autonomous learning skills for future training in the field.

This course covers the fundamentals of natural and medical image analysis and processing. The course is devoted to both the study of theoretical aspects and the scaling and application of algorithms for image analysis. The course program comprises the following topics.

Introduction.

• Definition and examples of areas of use of image processing. 
• Fundamental steps in image processing.
• Components of an image processing system.
• Light and the electromagnetic spectrum.
• Image acquisition and representation.
• Spatial and radiometric resolution.
• An overview of mathematical tools used in image processing.
• Compression formats for images and video.

Digital image processing in the spatial domain

• Brightness transformations and spatial filtering with related examples.
• Histogram processing and matching.
• Local histogram processing.
• Fundamentals of spatial filtering.
• Spatial smoothing and sharpening filters.
• Combination of space-based enhancement methods.

Digital image processing in the frequency domain.

• The Fourier transform of functions of a continuous variable.
• Sampling and the Fourier transform of sampled functions.
• Extension to functions of two variables.
• The 2D discrete Fourier transform and some of its properties.
• Fundamentals of filtering in the frequency domain.
• Smoothing and sharpening of images using filters in the frequency domain.

Image enhancement and restroration

• Periodic noise reduction using frequency domain filtering.
• Linear and non-position-dependent degradation.
• Inverse filtering. Reconstruction of images from projections.

Processing of color images.

• Color fundamentals.
• Processing of pseudo-color images.
• Transformations of colors.
• Segmentation of color images.

Low level vision

• Recognition of points, lines and boundaries of regions.
• Canny's edge detector.
• Thresholding.
• SIFT descriptors.

Advanced computer vision.

• Introduction to computer vision with statistical methods and neural networks.
• Discriminative and generative models, mixture of Gaussians, Factor Analysis.
• Fully connected artificial neural networks.
• Convolutional and fully convolutional networks for image classification.
• Deep architectures for image classification and related training methods.
• Neural networks for natural and medical image segmentation.
• Neural networks for localization and classifcation of objects in natural and medical images.
• Generative models.

Lectures take place both in traditional mode and through the use of multimedia tools. The student is invited to interact with the lecturer, and some experimental activities are aimed at deepening understanding of the topics presented in class.