FeaturesOverview

Image Features

• AutoColorCorrelogram (since v1.2.0) Based on [Image Indexing Using Color Correlograms](http://doi.ieeecomputersociety.org/10.1109/CVPR.1997.609412.
• CEDD
• Color Histogram
• FCTH
• Fuzzy Histogram
• FuzzyOpponentHistogram
• Gabor
• Haralick Texture Features (Wikipedia, Google Scholar)
• Histograms:
  • Gray-Histogram
  • RGB Histogram, including possibility to only extract Red, Green or Blue histograms
  • HSB Histogram, including possibility to only extract Hue, Saturation or Brightness histograms
• JCD
• LumnanceLayout
• MPEG7ColorLayout
• MPEG7EdgeHistogram
• Statistical Moments. The first 4 statistical moments: Mean, Standard Deviation, Skewneess and Kurtosis
• OpponentHistorgam
• PHoG (Pyramid Histograms of Oriented Gradients) Based on Representing Shape with a Spatial Pyramid Kernel.
• ReferenceColorSimilarity
• SURF
• Sift Wrapper for the SIFT binary. The binary can be obtained here
• Tamura
• Thumbnail

Some Features support masking. Masking can be used to extract features only from a certain part of the input image. A list of the feature descriptor capabilities can be obtained by java -cp JFeatureLib-x.x.x-jar-with-dependencies.jar de.lmu.ifi.dbs.jfeaturelib.utils.Extractor --list-capabilities

AutoColorCorrelogram     : [DOES_8G, DOES_8C, DOES_16, DOES_32, DOES_RGB]
CEDD                     : [NoChanges, DOES_8G, DOES_8C, DOES_32, DOES_RGB]
ColorHistogram           : [Masking, NoChanges, DOES_8C, DOES_16, DOES_32, DOES_RGB]
FCTH                     : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
FuzzyHistogram           : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
FuzzyOpponentHistogram   : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Gabor                    : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Haralick                 : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Histogram                : [Masking, NoChanges, DOES_8G, DOES_8C, DOES_RGB]
JCD                      : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
LumnanceLayout           : [DOES_8G]
MPEG7ColorLayout         : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
MPEG7EdgeHistogram       : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Moments                  : [DOES_8G, DOES_8C, DOES_16, DOES_32, DOES_RGB]
OpponentHistogram        : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
PHOG                     : [Masking, DOES_8G, DOES_8C, DOES_16, DOES_32]
ReferenceColorSimilarity : [Masking, NoChanges, DOES_8C, DOES_16, DOES_32, DOES_RGB]
SURF                     : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Sift                     : [DOES_8G]
Tamura                   : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]
Thumbnail                : [NoChanges, DOES_8G, DOES_8C, DOES_RGB]

Several implementations of features are origininally located in the LIRE-Project. JFeatureLib provides wrapper methods/classes so that these features can be extracted like others, too.

Point Detectors

• Harris Corner Detector
• Moravec
• Trajkovic And Hedley
• FAST - (paper)

Edge Detectors

• Sobel-Operator (in edgeDetector.Kernel)
• Scharr-Operator (in edgeDetector.Kernel)
• Prewitt-Operator (in edgeDetector.Kernel)
• Kirsch-Operator
• Canny
• Marr-Hildreth (paper)
• SUSAN
• Roberts-Operator
• Laplace-Filter (ImageJ Plugin)
• Derivative of Gaussian (DroG)

Shape Features

(keep in mind that those only work on binarized images, see issue 15)

• AdaptiveGridResolution
• CentroidBoundaryDistance
• CentroidFeature
• Compactness
• Eccentricity
• Extremal Points
• PolygonEvolution
• Profiles
• SquareModelShapeMatrix (since v1.1.0) also known as GridDescriptor. Based on Invariant shape description and measure of object similarity

Sources for other Features/Detectors

Several other features and point detectors are described and evaluated in the following publications:

• Hough transformation

• HIP Histogram Intensity Patches

• MPEG7

• Evaluation of Interest Point Detectors and Feature Descriptors for Visual Tracking

• Evaluation of Interest Point Detectors

• http://users.ecs.soton.ac.uk/msn/book/new_demo/

• CenSurE

• Histogram Intersection, Histogram Backprojection

• OpenCV

• http://koen.me/research/colordescriptors/readme

• http://www.semanticmetadata.net/features/

  • ColorLayout (measures color distribution in an image)
  • ScalableColor (basically a color histogram)
  • EdgeHistogram (measures edginess)
  • DominantColor (measure dominant colors)

• http://svg.dmi.unict.it/iplab/imagej/index.htm GPL licenced for jFeatureLib

• Surface Roughness (issue 1) ImageJ Plugin and description

• Low-dimensional and comprehensive color texture description

• Interest Operators

Shape descriptors

• Zernike moments
• Shape Survey
• another Shape Survey

Edge Detectors:

• http://en.wikipedia.org/wiki/Edge_detection / http://de.wikipedia.org/wiki/Kantendetektion include the links to
• Kontrastverstärker