Computer-Vision

Exercises for 263-5902-00L Computer Vision at ETH HS2019

1 Camera Calibration

• Data normalization
• Direct Linear Transform
• Gold Standard Algorithm

2 Feature Extraction

• Image gradients
• Local auto-correlation matrix
• Harris response function
• SSD one-way nearest neighbors matching
• Mutual nearest neighbors and ratio test

3 Particle Filter

• Monte Carlo Localization
• Markovian Localization problem using probability distribution
• Resampling of particles according their a posteriori proability weight after having a observation

4 Model Fitting

• RANSAC algorithm for line fitting
• Fundamental matrix estimation
• Eight-Point algorithm
• Feature extraction and matching
• adaptive version of RANSAC

5 Image Segmentation

• Lab color space
• Mean-shift Segmentation
• EM Segmentation

6 Stereo Matching

• Rectification
• Disparity computation using winner-takes-all stereo and SSD or SAD
• Graph cut
• Generation and visualization of textured 3D model using MeshLab

7 Structure from Motion

• Combination of all learned methods so far
• Feature extraction
• Fundamental matrix estimation using RANSAC and 8-point algorithm
• Triangulation
• Projection matrix estimation using RANSAC and DLT
• Visualization of camera positions and matched points

8 Shape Context

• Shape context descriptors
• Cost matrix computation
• Hungarian algorithm to find best matches
• Minimum distance sampling
• Thin plate spine