Images are everywhere. We live in a world full of audiovisual stimuli, and we are constantly bombarded with images and videos. If that wasn't enough, almost every one of us carries a smartphone in our pocket, and we're constantly taking pictures and recording videos of what's around us. In addition, digital image processing techniques are used in many fields, such as medicine, astronomy, industry, or biological sciences.
This course is an introduction to the broad area of digital image processing, providing students with fundamental knowledge in this field.
The course begins by presenting fundamental concepts for image analysis. In addition, techniques for image modification, enhancement and restoration are studied, emphasizing linear and non-linear filtering techniques. The problem of image segmentation is also studied. The final part of the course provides an introduction to computer vision, studying how machine learning and deep learning techniques are applied to solve tasks such as image classification and segmentation.
In addition, the course has a strong practical focus. The classes will be of a theoretical-practical nature, and students will need to attend with their laptops with Matlab installed. In addition, practices will be carried out that allow students to apply the contents of the subject in real problems of digital image processing.
Titular Professors
Digital Signal Processing is recommended.
Learning Outcomes of this subject are:
- LO.1 Basic general knowledge of digital image processing.
- LO.2 Ability to apply image processing knowledge to practice.
Unit 1. Introduction to digital image processing
1. The ubiquity of images
2. Types of images and applications of digital image processing
3. The human visual system
4. Acquisition of digital images
5. Basic image operations
6. Histogram of an image
7. Color spaces
Unit 2. Enhancing and restoring images
1. The importance of image enhancement and restoration
2. Noise reduction in images
3. Mathematical morphology
4. Contrast enhancement
5. Image sharpening
Unit 3. Image segmentation
1. The importance of image segmentation
2. Segmentation based on discontinuity
3. Segmentation based on similarity
4. Segmentation homogeneity criteria
Unit 4. Introduction to computer vision
1. The importance of image classification
2. Supervised learning
3. Classification based on decision theory
4. Classification based on local features
5. The bag-of-words approach.
6. The deep learning approach.
The course is taught in 2 weekly lessons, one lasting 100 minutes and the other one, 50.
The usual dynamics of each class will consist of a combination of theoretical explanations always followed by theoretical and/or practical exercises that exemplify what has just been explained. Applied methodologies: master class (MD0), problems and exercises class (MD1), lab practice (MD2).
Additionally, the eStudy provides resources for the student to carry out self-learning complementary practical activities. Applied methodology: self-paced learning (MD5).
Finally, in order to achieve an applied view of the concepts presented in class, two practical exercises in group using the Matlab software will be undertaken. Applied methodology: challenge-based learning (MD11).
To evaluate whether the student has achieved the course objectives to an adequate degree, the following assessments are used: Exams, In?class tests, Individual and group assignments.
The following will be assessed:
- Conceptual understanding of image processing: ability to explain the basic principles of image formation, representation and transformation, and clear understanding of the fundamentals of enhancement, restoration and segmentation.
- Correct application of enhancement and restoration techniques: appropriate selection of filtering, contrast and morphology methods depending on the objective, and precise application of procedures and coherent interpretation of the results.
- Analysis and implementation of segmentation methods: correct use of discontinuity or similarity-based techniques to separate regions of interest, ability to assess the quality of the segmentation obtained and justify the decisions.
- Integration of basic computer vision concepts: understanding of the essential elements of image classification and supervised approaches, ability to describe when it is appropriate to use classical or deep learning-based methods.
- Development of real-world problem-oriented practices: ability to solve applied problems integrating the techniques learned using Matlab, quality of the presentation of the results: clarity, structure and rigorous interpretation.
- Slides for lectures
- Anil K. Jain, `Fundamentals of digital image processing´, Prentice Hall, 1989
- Gonzalo Pajares, Jesús M. de la Cruz, `Visión por computador´, Ra-Ma
- Rafael C. Gonzalez, Richard E. Woods `Digital Image Processing´, Addison Wesley
- Arturo de la Escalera. Visión por computador. Prentice Hall 2001.
- Jain, Ramesh, Kasturi, Schunk, Brian. Machine Vision. MacGraw Hill, 1995.