||Seminar Advanced Digital Image Processing
||Dr. E.A. Hendriks
||Prof.dr.ir. L.J. van Vliet
|Contact Hours / Week x/x/x/x:
|Expected prior knowledge:
||signal processing (ET2560IN), image processing (TI2715-B), linear algebra (WI1530IN, WI1540IN), stochastic processes (ET3502 or ET3260IN).
The course will start with a brief review of basic image processing principles as discussed in TI2715-B.
||Image restoration (inverse filtering, Wiener filtering, geometric transformation), advanced morphological image processing and extension to grey-scale images, data-driven image segmentation (boundary detection, region-based segmentation, watersheds), model-based image segmentation (Hough transform, template matching, deformable templates, active contours), representation and description of image objects, image features (structure tensor, local shape), motion estimation (optical flow, feature-based techniques)
||General learning outcomes:
The student has insight into state of the art algorithms for image processing including Multi-Resolution Image Processing, Morphological Image Processing, Image Features Representation/Description, Motion Estimation and Optic Flow, Image Restoration, Image Segmentation and 3D Computer Vision. The student is able to read, discuss, summarize and comment on scientific journal and conference papers in this area.
Specific learning outcomes:
1. Multi-resolution Image Processing:
Gaussian scale space, windowed Fourier transform, Gabor filters, multi-resolution systems (pyramids, subband coding and Haar transform), multi-resolution expansions (scaling functions and wavelet functions), wavelet Transforms (Wave series expansion, Discrete Wavelet Transform (DWT), Continuous Wavelet Transform (CWT), Fast Wavelet Transform (FWT))
The student is able to motivate the use of space-frequency representations, analyze the behavior of space-frequency techniques, explain the principles behind, classify and evaluate multi-resolution techniques.
2. Morpological Image Processing:
Definitions of gray-scale morphology: erosion, dilation, opening, closing; Application of gray-scale morphology: smoothing, gradient, second derivatives (top hat), morphological sieves (granulometry).
The student is able to apply, recognize the priciples and analyze (a sequence of) morphological operations for noise suppression, edge detection, and sharpening.
3. Image Feature Representation and Description:
Measurement principles: accuracy vs. precision ; Size measurements: area and length (perimeter); Shape descriptors of the object outline: form factor, sphericity, eccentricity, curvature signature, bending energy, Fourier descriptors, convex hull, topology; Shape descriptors of the gray-scale object: moments, PCA, intensity and density; Structure tensor in 2D and 3D: Harris Stephens corner detector, isophote curvature.
The student is able to comprehend and explain the properties of measurements in digitized images, combine measurement principles to solve a new problem, comprehend the structure tensor in various notations and apply it in measurement procedures.
4. Motion and optic flow:
Motion is strcuture in spatio-temporal images; Two frame registration: Taylor expansion method; Multi-frame registration: Optic flow. Applications of image registration.
The student is able to explain the properties of image registration and optic flow and comprehend the aperture problem in optic flow.
5. Image Restoration:
Noise filtering, Wiener filtering, Inverse filtering, Geometric transformation, Grey value interpolation
The student is able to discuss the use of linear and non-linear noise filters, explain the use of inverse filters and problems of inverse filtering in the case of noise, describe (the use of) a Wiener filter and apply geometric transformations and bi-linear grey value interpolation
6. Image Segmentation:
Thresholding, edge and contour detection, data-driven and model-driven image segmentation, edge tracking
The student is able to discuss isodata thresholding, optimal thresholding, multimodal thresholding and adaptive thresholding techniques, apply Gaussian derivative filters and difference based filters for calculation of egde point candidates, explain the trade off between localization and detection of edges, discuss split and merge techniques and edge tracking techniques. The student has insight into model-based image segmentation (object detection) approaches like template matching, Hough Transform, Deformable Template matching, Active Contours and Active Shape models and is able to formulate how shape information and image intensity information can be incorporated into these approaches.
||Matlab and dipimage toolbox
|Literature and Study Materials:
||Book "Digital Image Processing", van R.C. Gonzalez en R.E. Woods, third edition, 2002, ISBN 9780131687288.
(Online) Book "Computer Vision, Algorithms and Applications", R. Szeliski, (http://szeliski.org/Book/). The online version is available for free.
We have used the Book “Introductory Techniques for 3-D Computer Vision”, E. Trucco and A. Verri, ISBN 0-13-261108-2 in the past.
Lecture notes “Fundamentals of Image Processing”
PDF-files of the lecture slides (see blackboard)
||written exam and assignment
||There will be a written examination in the exam period after the first semester. The assessment of the assignment will take place at the end of the first semester or in the exam period after the first semester.
|Permitted Materials during Tests:
||Books, print-out of pdf files of the lecture slides and lecture notes are not permitted during the written examination