Planeamento
Aulas Teóricas
T1 - Introduction to the course.
Introduction to the course. Introduction to image processing and computer vision, assessment methods and bibliography. Introduction to image processing and computer vision. Colour spaces. Conversion between colour spaces. Representation of images in grayscale.
T2 - Spatial linear and nonlinear filters for gray level images.
Spatial linear and nonlinear filters for gray level images. Types of image noise. Mechanics of spatial filtering. Smoothing spatial filters: type of filters.. Binary mathematical morphology Linear filters: Sharpening spatial filters: image gradient, smoothed derivative, Image Laplacian, Laplacian of Gaussian filter, high-boost filtering. Brief introduction to nonlinear filters: median filter; morphological gray level filters.
T3 - Point Processing of images.
Point Processing of images. Image binarization. Introduction to linear filter for grayscale images. Histogram based operations. Histogram Equalization. Automatic and manual image binarization. Otsu's method. Introduction to binary mathematical morphology, mathematical basis and main operations. Introduction to image segmentation Boundary extraction; Region filling; Hit-or-Miss transformation; Skeletonization; Morphological reconstruction; Convex Hull; Euler Number; Ultimate erosion.
T4 - Introduction to image segmentation.
Introduction to image segmentation. Grouping and segmentation. Methods of segmentation. Edges detection. The Canny edge detector. Introduction to shape descriptors. The Hough transform to detect lines, circles and general shapes. Corner detection. The Harris detector. Region based methods. Region Splitting and Merging. Superpixels. The SLIC algorithm. Introduction to shape descriptors. Why image descriptors. Types of invariance. Shape as a region: area, Euler number, eccentricity, geometric moments, invariant moments. Descriptors based on the shape skeleton.
T5 - Shape descriptors (conclusion)
Shape descriptors (conclusion). Image Description Based on Texture. Contour based descriptors: perimeter, shape signatures, chain code, Fourier descriptors. Definition of texture. Texture primitives. Texture descriptors: statistical, structural and spectral approach. The auto-correlation. Frequency Descriptors. Co-Occurrence matrices. Law’s texture energy measures.
T6 - Patches descriptors from image interest points
Patches descriptors from image interest points. Methods to find interest points. SIFT algorithm overview: evaluate keypoints and build keypoints descriptors. Others local features detectors and descriptors: the SURF algorithm; binary descriptors: BRIEF, ORB, BRISK, FREAK.
T7 - Intelligent vision systems
Intelligent vision systems. Intelligent vision systems for segmentation and classification. Basis of classification and learning. Distance and classification. Neural Networks and Support Vector Machines.
T8 - Deep learning
Deep learning. Basis of deep learning. Image datasets and data partition. Data augmentation. Pretrained networks. Transfer learning.
T9 - Camera models and calibration
Camera models and calibration. Pinhole camera model. Lenses. Depth of field. Field of view. Geometric distortion and chromatic aberration. Perspective imaging. Homogeneous coordinates. Perspective projection. Camera calibration. Extrinsic and intrinsic parameters.
T10 - Stereo vision
Stereo vision. Basic stereo geometry. Disparity map. Epipolar geometry. Image rectification. Stereo correspondence. Occlusion problem. Dynamic programming formulation.
T11 - Projective geometry
Projective geometry. Homogeneous coordinates and projective geometry. Representation of a line, duality and ideal points. Projective similarity. Line and point duality. Ideal points and lines. 2D transformations in projective space. Image reprojection - homography. Image warping. The epipolar geometry. Stereo geometry with calibrated cameras. The Essential matrix E. The fundamental matrix F.
Revisions (Kahoots) and project support
Revisions (Kahoots) and project support.
Project support
Project support.
Aulas Laboratoriais
Laboratory 1 - Colour Spaces and Histograms
- Familiarization with image loading and visualization tools in MATLAB and Simulink
- Experiment different colour spaces
- Learn about the histogram of a grayscale image
Laboratory 2 - Spatial Filters
- Learn how different noise types affect a grayscale image
- Explore several algorithms in spatial filtering for image smoothing
- Understand the advantages and limitations of particular algorithms
Laboratory 3 - Point Processing and Binary Morphology
- Explore the common point processing techniques and study how these impact their histogram
- Learn how to threshold a grayscale image using a suitable method
- Familiarization with different morphological operations
Laboratory 4 - Image Segmentation
- Familiarization with common edge, corner and circle detection in images
- Study the application spatial filtering on image segmentation
Laboratory 5 - Image Description Based on Keypoints
- Study the available feature detection methods
- Learn how to use feature extraction and matching for classification
Laboratory 6 - Image Description Based on Shape
- Learn to extract region properties of binary images
- Explore and analyse online shape classification
- Familiarization with superpixels and how these can describe images
Project 1 Support
- Questions and doubt clarification about Project 1 of the course
Laboratory 7 - Intelligent Vision Systems
- Study and apply different classification models
- Learn to classify images based on classification models
- Analyse the obtained results using adequate metrics
Laboratory 8 - Deep Learning for Image Recognition
- Use a pre-trained neural network for generic image classification
- Create a simple neural network for digit classification
- Analyse the obtained results
Laboratory 9 - Camera Calibration
- Study the mechanics of calibrating a stereo camera setup
- Comprehend how to analyse the image using the intrinsic and extrinsic parameters
Laboratory 10 - Stereo Vision
- Comprehend how the intrinsic and extrinsic parameters of a stereo rig can be used to extract information from images
- Understand how key concepts of Projective Geometry are interconnected
MAP 2 and Project 2 Support
- MAP 2 of the course
- Questions and doubt clarification about Project 2 of the course