stereo_matching_flashcards
(20 cards)
What is the goal of stereo matching?
To find correspondences between two images to reconstruct 3D scene geometry.
What is image rectification in stereo vision?
The process of reprojecting image planes so that corresponding points lie on the same horizontal scanline.
What is disparity in stereo vision?
The horizontal shift between matching pixels in rectified stereo image pairs.
What is the Sum of Absolute Differences (SAD)?
A window-based matching cost function used to compare pixel intensities.
What are common issues with local stereo matching?
Photometric variations, specularities, textureless regions, repetitive structures, and occlusions.
What is scanline stereo?
A method that matches pixels along each scanline independently using dynamic programming.
What are the three cases in scanline stereo correspondence search?
Sequential match, occluded, and disoccluded.
What is dynamic programming used for in stereo?
To find the optimal correspondence path along a scanline under ordering constraints.
What is the limitation of scanline stereo?
It produces streaking artifacts and doesn’t enforce global consistency across scanlines.
What is Semi-Global Matching (SGM)?
An algorithm that combines local and global stereo methods for dense, accurate matching.
Why is dense matching preferred over sparse?
It provides detailed geometric information and more complete 3D reconstructions.
What does a 7x7 window mean in window-based matching?
A matching block of 49 pixels used to compare regions between images.
What is the main motivation for dense stereo matching?
To obtain more points for accurate and detailed 3D surface reconstruction.
What is the role of RANSAC in stereo vision?
To identify and reject outliers in feature correspondences before dense matching.
Why is the correspondence problem challenging?
Descriptor similarity alone can result in incorrect matches due to ambiguities.
What is the ‘Stereo Normal Case’?
When images are on the same plane, have identical orientation, and baseline is along x-axis.
What is the effect of window size in matching?
Larger windows reduce ambiguity but increase smoothing and may blur fine details.
What is the ordering constraint in stereo?
The assumption that points appear in the same order along scanlines in both images.
Who proposed Semi-Global Matching and when?
Heiko Hirschmüller in 2005 and 2008.
What benchmark is commonly used to evaluate stereo algorithms?
Middlebury stereo benchmark (e.g., Tsukuba dataset).
