AbsDiffTransform

Take the absolute difference of two matrices.

AbsTransform

Computes the absolute value of each element.

AdaptiveThresholdTransform

Wraps OpenCV's adaptive thresholding.

AffineTransform

Performs a two or three point registration.

AndTransform

Logical AND of two matrices.

ApplyMaskTransform

Applies a mask from the metadata.

BayesianQuantizationTransform

Quantize into a space where L1 distance approximates log-likelihood.

  • file: imgproc/bayesianquantization.cpp
  • inherits: Transform
  • author(s): Josh Klontz
  • properties: None

BinarizeTransform

Approximate floats as signed bit.

BlendTransform

Alpha-blend matrices

BlurTransform

Gaussian blur

BuildScalesTransform

DOCUMENT ME

ByRowTransform

Turns each row into its own matrix.

CannyTransform

Wrapper to OpenCV Canny edge detector

CatColsTransform

Concatenates all input matrices by column into a single matrix. Use after a fork to concatenate two feature matrices by column.

CatRowsTransform

Concatenates all input matrices by row into a single matrix. All matricies must have the same column counts.

CatTransform

Concatenates all input matrices into a single matrix.

CenterTransform

Normalize each dimension based on training data.

ContrastEqTransform

Perform contrast equalization

  • file: imgproc/contrasteq.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • read:

    1. Xiaoyang Tan; Triggs, B.;
      "Enhanced Local Texture Feature Sets for Face Recognition Under Difficult Lighting Conditions," Image Processing, IEEE Transactions on , vol.19, no.6, pp.1635-1650, June 2010

  • properties: None

ConvexHullTransform

Set the template's label to the area of the largest convex hull.

CropBlackTransform

Crop out black borders

CropFromLandmarksTransform

Crops around the landmarks numbers provided.

CropFromMaskTransform

Crops image based on mask metadata

CropSquareTransform

Trim the image so the width and the height are the same size.

CropTransform

Crops about the specified region of interest.

CryptographicHashTransform

Wraps QCryptographicHash

CustomSIFTTransform

Specialize wrapper OpenCV SIFT wrapper

CvtFloatTransform

Convert to floating point format.

CvtTransform

Colorspace conversion.

CvtUCharTransform

Convert to uchar format

DiscardAlphaTransform

Drop the alpha channel (if exists).

DivTransform

Enforce a multiple of n columns.

DoGTransform

Difference of gaussians

DownsampleTransform

Downsample the rows and columns of a matrix.

DupTransform

Duplicates the Template data.

EnsureChannelsTransform

Enforce the matrix has a certain number of channels by adding or removing channels.

EqualizeHistTransform

Histogram equalization

FillContoursTransform

Fills contours with white pixels.

FlipTransform

Flips the image about an axis.

FloodTransform

Fill black pixels with the specified color.

GaborJetTransform

A vector of gabor wavelets applied at a point.

GaborTransform

Implements a Gabor Filter

GammaTransform

Gamma correction

GradientMaskTransform

Masks image according to pixel change.

GradientTransform

Computes magnitude and/or angle of image.

GroupTransform

Group all input matrices into a single matrix.

Similar to CatTransfrom but groups every size adjacent matricies.

HeatmapTransform

Applies a colormap to src.

HistBinTransform

Quantizes the values into bins.

HistEqQuantizationTransform

Approximate floats as uchar with different scalings for each dimension.

HistTransform

Histograms the matrix

HoGDescriptorTransform

OpenCV HOGDescriptor wrapper

HuMomentsTransform

Computes the seven Hu moments of a shape.

IfTransform

Document Me!

InpaintTransform

Wraps OpenCV inpainting

IntegralHistTransform

An integral histogram

IntegralSamplerTransform

Sliding window feature extraction from a multi-channel integral image.

IntegralTransform

Computes integral image.

KernelHashTransform

Kernel hash

KeyPointDescriptorTransform

Wraps OpenCV Key Point Descriptor

LBPTransform

Convert the image into a feature vector using Local Binary Patterns

  • file: imgproc/lbp.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • read:

    1. Ahonen, T.; Hadid, A.; Pietikainen, M.;
      "Face Description with Local Binary Patterns: Application to Face Recognition" Pattern Analysis and Machine Intelligence, IEEE Transactions, vol.28, no.12, pp.2037-2041, Dec. 2006

  • properties: None

LTPTransform

DOCUMENT ME

  • file: imgproc/ltp.cpp
  • inherits: UntrainableTransform
  • author(s): Brendan Klare, Josh Klontz

  • read:

    1. Tan, Xiaoyang, and Bill Triggs.
      "Enhanced local texture feature sets for face recognition under difficult lighting conditions." Analysis and Modeling of Faces and Gestures. Springer Berlin Heidelberg, 2007. 168-182.

  • properties: None

LargestConvexAreaTransform

Set the template's label to the area of the largest convex hull.

LimitSizeTransform

Limit the size of the template

MAddTransform

dst = a*src+b

MaskTransform

Applies an eliptical mask

MatStatsTransform

Statistics

MeanFillTransform

Fill 0 pixels with the mean of non-0 pixels.

MeanTransform

Computes the mean of a set of templates.

Suitable for visualization only as it sets every projected template to the mean template.

MergeTransform

Wraps OpenCV merge

MorphTransform

Morphological operator

NLMeansDenoisingTransform

Wraps OpenCV Non-Local Means Denoising

NormalizeTransform

Normalize matrix to unit length

  • file: imgproc/normalize.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • properties:

    Property Type Description
    NormType enum Values are:
    • NORM_INF
    • NORM_L1
    • NORM_L2
    • NORM_MINMAX
    ByRow bool If true normalize each row independently otherwise normalize the entire matrix.
    alpha int Lower bound if using NORM_MINMAX. Value to normalize to otherwise.
    beta int Upper bound if using NORM_MINMAX. Not used otherwise.
    squareRoot bool If true compute the signed square root of the output after normalization.

OrigLinearRegressionTransform

Prediction with magic numbers from jmp; must get input as blue;green;red

PackTransform

Compress two uchar into one uchar.

PowTransform

Raise each element to the specified power.

ProductQuantizationDistance

Distance in a product quantized space

  • file: imgproc/productquantization.cpp
  • inherits: UntrainableDistance
  • author(s): Josh Klontz

  • read:

    1. Jegou, Herve, Matthijs Douze, and Cordelia Schmid.
      "Product quantization for nearest neighbor search." Pattern Analysis and Machine Intelligence, IEEE Transactions on 33.1 (2011): 117-128

  • properties: None

ProductQuantizationTransform

Product quantization

  • file: imgproc/productquantization.cpp
  • inherits: Transform
  • author(s): Josh Klontz

  • read:

    1. Jegou, Herve, Matthijs Douze, and Cordelia Schmid.
      "Product quantization for nearest neighbor search." Pattern Analysis and Machine Intelligence, IEEE Transactions on 33.1 (2011): 117-128

  • properties: None

QuantizeTransform

Approximate floats as uchar.

RGTransform

Normalized RG color space.

ROIFromPtsTransform

Crops the rectangular regions of interest from given points and sizes.

ROITransform

Crops the rectangular regions of interest.

  • file: imgproc/roi.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • properties:

    Property Type Description
    propName QString Optional property name for a rectangle in metadata. If no propName is given the transform will use rects stored in the file.rects field or build a rectangle using "X", "Y", "Width", and "Height" fields if they exist.
    copyOnCrop bool If true make a clone of each crop before appending the crop to dst. This guarantees that the crops will be continuous in memory, which is an occasionally useful property. Default is false.

RankTransform

Converts each element to its rank-ordered value.

RectRegionsTransform

Subdivide matrix into rectangular subregions.

RecursiveIntegralSamplerTransform

Construct Template in a recursive decent manner.

  • file: imgproc/recursiveintegralsampler.cpp
  • inherits: Transform
  • author(s): Josh Klontz
  • properties: None

RedLinearRegressionTransform

Prediction using only the red wavelength; magic numbers from jmp

ReshapeTransform

Reshape each matrix to the specified number of rows.

ResizeFilterTransform

Resize the template depending on its metadata

  • file: imgproc/resizefilter.cpp
  • inherits: UntrainableTransform
  • author(s): [Jordan Cheney][JordanCheney]
  • properties: None

ResizeTransform

Resize the template

  • file: imgproc/resize.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • properties:

    Property Type Description
    method enum Resize method. Good options are:
    • Area should be used for shrinking an image
    • Cubic for slow but accurate enlargment
    • Bilin for fast enlargement
    preserveAspect bool If true, the image will be sized per specification, but a border will be applied to preserve aspect ratio.

RevertAffineTransform

Designed for use after eye detection + Stasm, this will revert the detected landmarks to the original coordinate space before affine alignment to the stasm mean shape. The storeAffine parameter must be set to true when calling AffineTransform before this.

RndAffineTransform

Perform a number of random transformations to the points in metadata as "Affine_0" and "Affine_1"

  • file: imgproc/rndaffine.cpp
  • inherits: UntrainableMetaTransform
  • author(s): Jordan Cheney

  • properties:

    Property Type Description
    numAffines int The number of independent random transformations to perform. The result of each transform is stored as its own template in the output TemplateList
    scaleFactor float Controls the magnitude of the random changes to the affine points
    maxAngle int the maximum angle between the original line between the two affine points and the new line between the points.

RndPointTransform

Generates a random landmark.

RndRegionTransform

Selects a random region.

RndRotateTransform

Randomly rotates an image in a specified range.

RndSubspaceTransform

Generates a random subspace.

RootNormTransform

dst=sqrt(norm_L1(src)) proposed as RootSIFT (see paper)

  • file: imgproc/rootnorm.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • read:

    1. Arandjelovic, Relja, and Andrew Zisserman.
      "Three things everyone should know to improve object retrieval." Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. IEEE, 2012.

  • properties: None

RowWiseMeanCenterTransform

Remove the row-wise training set average.

SIFTDescriptorTransform

Specialize wrapper OpenCV SIFT wrapper

SampleFromMaskTransform

Samples pixels from a mask.

ScaleMatTransform

Scales the mat values by provided factor

ScaleTransform

Scales using the given factor

SkinMaskTransform

Make a mask over skin in an image

SlidingWindowTransform

Sliding Window Framework for object detection. Performs an exhaustive search of an image by sliding a window of a given size around the image and then resizing the image and repeating until terminating conditions are met.

  • file: imgproc/slidingwindow.cpp
  • inherits: MetaTransform
  • author(s): Jordan Cheney, Scott Klum

  • properties:

    Property Type Description
    classifier Classifier* The classifier that determines if a given window is a positive or negative sample. The size of the window is determined using the classifiers windowSize method.
    minSize int The smallest sized object to detect in pixels
    maxSize int The largest sized object to detect in pixels. A negative value will set maxSize == image size
    scaleFactor float The factor to scale the image by during each resize.
    minNeighbors int Parameter for non-maximum supression
    confidenceThreshold float A threshold for positive detections. Positive detections returned by the classifier that have confidences below this threshold are considered negative detections.
    eps float Parameter for non-maximum supression

SobelTransform

Approximates the gradient in an image using sobel operator.

SplitChannelsTransform

Split a multi-channel matrix into several single-channel matrices.

SplitRowsTransform

Split a multi-row matrix into several single-row or smaller multi-row matrices.

  • file: imgproc/split.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • properties:

    Property Type Description
    step int The numbers of rows to include in each output matrix.

SubdivideTransform

Divide the matrix into 4 smaller matricies of equal size.

SubtractTransform

Subtract two matrices.

SynthesizePointsTransform

Synthesize additional points via triangulation.

ThresholdTransform

Wraps OpenCV's thresholding.

TransposeTransform

Get the transpose of the Template matrix

TrimTransform

Trims a percentage of width and height from the border of a matrix.

WatershedSegmentationTransform

Applies watershed segmentation.