| Topic |
Sub-topic |
Algorithm
/Author |
Level |
Concept |
Useful
links |
| Development |
Software
tools |
Matlab |
* |
widely used, fast development, slow run time |
homepage |
| openCV |
** |
fast run time, medium documentation |
documentation |
| CUDA |
*** |
very-fast, difficult to use |
homepage |
| Books |
Image processing |
Gonzales &
Woods |
* |
great for beginners and to review the material |
amazon |
| signal processing | DSP guide | * | great for understand signals and FFT | free |
| computer vision | Szeliski | ** | most up-to-date. few explanations, many good references | free |
| 3D reconstruction |
Hartley &
Zisserman |
** |
in-depth understanding of 3D concepts |
amazon |
| Usefull links | matlab code | Matlab Central | * | large community, open source | link |
| Peter Kovesi | * | many good implementations and demos | link |
| papers | VisionBib | * | huge list of papers | link |
| CVpapers | * | papers from the most important conferences | link |
| General
concepts |
Math |
FFT |
* |
|
DSP book |
| Wavelets |
** |
|
|
| Gabor filters |
** |
|
wiki, intro |
| Kalman filter |
* |
|
wiki,
intro |
| Markov Random Fields |
** |
|
wiki |
Feature detection
|
Edge detection |
Sobel, Scharr, Roberts, Prewitt |
* |
Fast 1st derivative
approximation |
|
| LoG |
* |
|
|
| Canny |
** |
Smooth, gradient, non-maximum suppression, hysteresis thrsholding |
wiki |
| phase congruency |
** |
location of matched phases correspond to edges |
description |
| Corner detection |
Hessian |
* |
|
|
| FAST | ** | | openCV |
| Line
detection |
Frangi |
** |
line's 2nd derivative should be one-directional and strong |
matlab |
| Hough transform |
* |
find lines in parameter-space |
wiki |
| Features for registration (Key points) | SIFT | ** | | |
| SURF | ** | | |
| Registration |
Basic |
Normalized Cross
Correlation |
* |
invariant for mean+gain differences |
|
| phase-shift-correlation |
** |
|
wiki |
| Optic flow |
Lucas-Kanade |
** |
|
wiki |
| Horn-Schunck |
** |
|
wiki |
| Noise removal |
Basic |
mean filter |
* |
good for gaussian noise |
|
| median filter |
* |
good for salt-and-pepper |
|
| Wiener filter |
** |
estimation of noise spectrum |
|
| Advanced |
Bilateral filter |
** |
averaged pixels must be similar in both location and value
|
papers and code |
| Non-local means |
** |
like bilateral, but use region-values instead of pixel-value. slow |
|
| Anistropic
(Perona-Malik) |
** |
preserve edges |
|
| Classification |
|
Maximum-Likelihood
(ML) |
* |
|
|
| Maximum Aposteriori
(MAP) |
* |
|
|
| Fisher |
* |
|
|
| Nearest-neighbor, FLANN |
** |
good if classes are highly mixed |
|
| adaboost |
** |
use several weak classifiers to create one strong classifier |
|
| Support Vector
Machines (SVM) |
** |
maximize the separation between classes |
|
| Neural Networks (NN) |
*** |
|
|
Clustering
| | k-means | * | parametric, known number of clusters, fast | wiki |
| mean-shift | * | non-parametric, unknown number of clusters, slow | tutorial |
| Segmentation | | graph cuts |
** |
|
wiki |
| Hierarchical
clustering |
** |
|
|
| grow cut | *** | interactive | buy |
| Dynamic
programming |
|
Vitterbi |
* |
|
|
| Dijkstra |
** |
|
|
| Active contour model (snakes) |
** |
|
|
| Parameter
esimation |
|
Gaussian Mixture
Model (GMM) |
** |
|
|
| RANSAC |
** |
Test correctness of random-samples of data |
wiki |
| Detection |
Face
detection |
Viola & Jones |
* |
|
pdf |
| Algorithms collection |
** |
|
link |
| Optimization |
|
Gradient descent
|
* |
|
|
| Simulated annealing |
** |
|
wiki |
| Image
enhancement |
Contrast
improvement |
Histogram
equalization |
* |
|
wiki |
| Gamma correction |
* |
|
wiki |
| Sharpenning |
unsharp masking |
* |
add the high-pass to the image |
wiki |
| Video analysis | boundary tracking | level sets | ** | | intro |
| fast marching | ** | | intro |
| 3D |
General topics |
Epipolar geometry |
* |
|
wiki |