Machine Vision
Machine vision from first principles: image formation, optical flow, shape from shading, edge and object detection, and photogrammetry.
Machine vision turns images into a symbolic description of the environment, for robots and other machines that interact with the world. This course works from the physics of image formation, asking where a point will image and how bright it will be, then inverts it: motion and optical flow, photometric stereo and shape from shading, edge and blob analysis, object detection and pose, photogrammetry and binocular stereo.
What You Will Achieve
- Model image formation, perspective projection, and the image irradiance equation
- Recover motion, time to contact, and optical flow from image sequences
- Reconstruct shape from shading and photometric stereo
- Determine object pose and camera orientation with photogrammetry and binocular stereo
Program Overview
Image Formation and Motion
How images form and what motion reveals.
- Image formation and perspective projection
- Motion field, time to contact, and focus of expansion
- Optical flow and the constant brightness assumption
- Vanishing points and camera calibration
Shape from Images
Recovering surfaces from brightness.
- Photometric stereo and noise gain
- Gradient space and the reflectance map
- Shape from shading
- Characteristic strip expansion and iterative solutions
Detection, Recognition and Orientation
Finding objects and where the camera is.
- Edge detection, blob analysis, and binary image processing
- Object detection, recognition, and pose determination
- Fast convolution and filtering
- Photogrammetry, rotation, binocular stereo, and bundle adjustment
Course Completion
Job Outlook
Machine vision is used in robotics, industrial inspection, autonomous vehicles, photogrammetry and mapping, and computer vision research.
Good for: Learners with strong math who want the physics-based foundations of machine vision, derived step by step.
Keep in mind: The materials are lecture notes, five homework sets and two take-home quizzes, with no videos; the final project was only for the graduate version.
Bottom line: A physics-first machine vision course that derives each method, including several used in industrial inspection systems.
Frequently Asked Questions
How does the course represent rotations?
Lectures cover how to represent rotation with unit quaternions and the space of rotations, then use them for absolute orientation in closed form, outliers and RANSAC.
What is the project for the graduate version?
An open-ended project applying the course to a vision problem, such as time to contact, optical flow, headlight tracking or vanishing points, using OpenCV, Android or MontiVision.
Are industrial methods discussed?
Yes. Several lectures cover patented methods for edge detection, object detection and alignment used in industrial inspection, including PatQuick and PatMAx.



There are no reviews yet.