CS413 15 CATS (7.5 ECTS) Term 1
Option - MEng CS and DM, MSc CS, MSc DA
CS118, CS131 or equivalent knowledge and experience
The module aims to teach the fundamentals of digital image processing, image and video analysis. In particular, it will present the mathematics and algorithms that underlie image analysis techniques such as filtering, edge detection, feature detection, tracking and 3D reconstruction. It will also present how these tools are used in algorithms for image and video segmentation, motion estimation, stereo reconstruction, video analysis, object detection and recognition.
The course will enable computer science undergraduates to apply their mathematical knowledge and understanding of algorithms to problems in image and video processing: from preprocessing, to quantitation, video compression and video interpretation, machine learning for image analysis. The methods have numerous applications e.g. in medicine, biology, robotics (computer vision), surveillance, security, biometrics, database searching, TV and entertainment.
At the end of the module the student should be able to:
- Understand and describe the fundamental principles of image and video analysis and have an idea of their application
- Perform written communication in coursework reports and examination. Use of Python and associated image/video analysis libraries for coding. Critically analyse multimedia applications
Assimilate mathematical and algorithmic material on image processing fundamentals and appropriately apply and program solutions to real problems in image and video analysis
- Understand and know how to apply state-of-art machine learning techniques (convolutional neural networks) to solving problems in image and video analysis
- Introduction to Human visual perception.
- Image sampling and quantization.
- Filtering by convolution and correlation: blurring, sharpening, edge detection.
- Colour models and contrast processing.
- Histograms and basic statistical models of images. Information and entropy.
- Discrete Fourier transforms and Discrete Cosine Transforms for filtering, coding, correlation and motion estimation.
- Image Pyramids for analysis and image compression.
- Image and video segmentation and texture models.
- Algorithms for boundary and region segmentation from images and video, including foreground detection
- Optical flow techniques for motion estimation and tracking
- Pinhole camera models and 3D reconstruction, including the principles image stitching and augmented reality
- Feature detection and object recognition in images
- Neural-networks for image analysis and end-to-end learning: from fully-connected networks to convolutional neural networks
- Programming image and video analysis methods in Python + associated libraries, OpenCV and Keras
- R.C. Gonzalez and R.E. Woods. Digital Image Processing. 3rd Edition. Addison Wesley, 2007.
- Mark S. Nixon and Alberto S. Aguado. Feature Extraction and Image Processing for Computer Vision. Academic Press. 2012
- Simon J. D. Prince. Computer Vision: Models, Learning and Inference. Cambridge University Press, 2013.
Two hour examination (70%), laboratories and coursework (30%)
30 one-hour lectures plus 8 two-hour workshops