Image and video analysis. Part 2

Lecture 1. Image segmentation

Texture representation and classification, Gabor filters, textons, Pb boundary detector.
Heuristic methods – region growing, split and merge, watershed.
Graph-based image segmentation.
Clustering-based methods – MeanShift, QuickShift.
Energy minimization methods and level sets - Snakes, TurboPixels.

Lecture 2. Introduction to graphical models

Random fields, factor graph, graphical model.
Markov random fields and Hammersley-Clifford theorem.
Energy minimization in graphical models - MinCut and a-expansion.

Lecture 3. Digital photomontage

Foreground extraction – GraphCut and GrabCut.
Image matting problem and approaches.
Image blending – feathering, Laplace pyramid, Poisson blending.
Image stitching with graph cuts.
Image inpainting and texture synthesis.

Lecture 4. Semantic image segmentation and video segmentation

Oversegmentation and superpixels, spatial support.
General pipeline of semantic image segmentation.
Graphical models for video volume segmentation.

Lecture 5. Part-based object detection and localization

Statistical shape models for feature detection, active shape models, active appearance models, local-constrained models.
Pictorial structures, application to facial feature detection and pose estimation.

Lecture 6. Face recognition

Face recognition problems – face attributes, face verification, identification, watch-list.
Features for face recognition - local binary patterns, biologically inspired features, feature learning, “similes”.
Application of face recognition methods.

Lecture 7. Deep learning

Introduction to neural networks.
Various hierarchical models.
Application of deep learning methods to image recognition problems.

Lecture 8. Camera model and multiple view geometry

Camera model – internal and external calibration, distortion, camera calibration.
Homography and its estimation.
Fundamental and essential matrices and their estimation.

Lecture 9. Structure from motion

Structure estimation – point triangulation.
Iterative structure from motion.
Bundle Adjustment.
3D reconstruction from community photo collections.

Lecture 10. Dense stereo

Basics of dense stereo - image rectification, triangulation, visual cues.
Local and global methods for dense stereo.
Segmentation-based methods.

Lecture 11. Multiple view stereo

Voxel-based methods.
Depth map fusion.
Patch-based multiview stereo.
Application - 3D reconstruction of human head.

Lecture 12. Single-view 3D reconstruction

Vanishing points and lines.
Interactive single-view reconstruction methods.
Joint image recognition and 3D reconstruction.
All lectures are accompanied by practical tasks for seminars and homework. Also students are expected to participate in one course project – image recognition competition.

Literature:

R.Szeliski “Computer vision: algorithms and applications”.
D.A. Forsyth, J. Ponce “Computer Vision: A Modern Approach (2nd Edition)”
R.Gonsalez, R. Woods, “Digital image processing”