Table of Contents
Pattern recognition - Basic
Mean-Square Regression
Objective
Very fast if you already have a list of pixels you know belong to one contour, and you want to check if it fits to a parametric shape.
Quick Def
References
Full Definition
Hough Transforms
Standard Hough Transform
Randomized Hough Transform
[Xu,Oja,Kultanen,1989]
Objective
Improve speed, resolution, low memory needs, infinite scale.
Quick Def
If n parameters, take n points and only accumulate one point.
References
Full Definition
Connective Randomized Hough Transform
[Kalvianen,Hirvonen]
Objective
Quick Def
References
=Full Definition=
Combinatorial Hough Transform
Objective
Quick Def
References
Full Definition
Adaptive Hough Transform
[Ilingworth,Kittler,1986]
Objective
Improve speed and resolution.
Quick Def
First time at low resolution, then second time at higher resolution where there are peaks.
References
Full Definition
Probabilistic Hough Transform
[Kiryati,Eldar,Bruckshtein,1990]
Objective
Improve speed.
Quick Def
Only process n% of the pixels.
References
Full Definition
Adaptive Probabilistic Hough Transform
Objective
Quick Def
References
Full Definition
Progressive Probabilistic Hough Transform
Objective
Quick Def
References
Full Definition
Hierarchical Hough Transform
[Princen,1989]
Objective
Quick Def
References
Full Definition
Sampling Hough Transform
Objective
Quick Def
References
Generalized Hough Transform
Objective
Quick Def
References
Full Definition
UpWrite method
References
Curvogram
Pattern recognition - Learning
Viola-Jones Detector
[Viola,Jones, 2001]
Quick Def
Cascade of boosted classifiers based on Haar-like features.
References
with Extended Set of Haar features
[Lienhart,Maydt, 2002]
Quick Def
Basic features are line features, edge features, and center surrounded features, all of them horizontal or vertical. The extended set also includes 45° rotated features, and an optional diagonal feature.
References
Stumps or CART trees
Quick Def
Stumps are single node trees, and CART (Classification And Regression Trees) more complex trees with a few nodes (4 or 5).
Rotation Invariant
References
Multiview
Parallel Cascades
Objective
Obvious way to deal with multiview.