Free online courseComputer Vision Fundamentals: Image Formation, Filtering and Machine Learning

Course Description

Computer Vision Fundamentals: Image Formation, Filtering and Machine Learning is a free online course in Technology and Programming that builds a solid foundation for understanding how images are created, processed, analyzed, and interpreted by computer systems.

You will start with the physics and geometry behind image formation, exploring the pinhole camera model, perspective projection in 2D and 3D, and the role of lenses in focus and exposure. The course also explains real-world imaging details such as sensors, color filter arrays, displays, common artifacts, lens distortion, noise, and compression effects that influence what visual data looks like before any algorithm touches it.

From there, you will move into image filtering with a clear, math-driven view of convolution, including 1D and 2D filtering, linear time-invariant systems, and separable filters. You will connect spatial and frequency perspectives through Fourier analysis and sampling concepts, then learn multiscale representations such as Gaussian and Laplacian pyramids and practical feature extraction ideas like edges, line detection, and histogram of oriented gradients.

The course then introduces core techniques for image analysis, including motion estimation and feature tracking, stereo vision for depth, epipolar constraints, and planar homography with applied examples. Finally, it transitions to image understanding through machine learning, covering supervised regression and classification as well as unsupervised methods. You will encounter key models and optimization ideas used in practice, along with widely used approaches such as logistic regression, LDA, ROC analysis, support vector machines, neural networks including convolutional architectures, clustering with k-means, expectation maximization, PCA and eigenfaces, and tSNE for visualization.

Designed for learners who want both intuition and rigor, this course connects camera models, signal processing, geometry, and learning-based methods into one coherent pathway toward modern computer vision.