dataStructures - lists the data structures used in the SaliencyToolbox.
DATA STRUCTURES USED IN THE SALIENCYTOOLBOX
Global variables
IS_INITIALIZED: flag that initializeGlobal was called.
IMG_EXTENSIONS: cell arrays with possible extensions for image files.
DEBUG_FID: file identifier for debugMsg output.
PD: path delimiter for your operating system.
BASE_DIR: base directory for data and image locations.
IMG_DIR: directory for images.
DATA_DIR: directory for data.
TMP_DIR: directory for temporary files.
See also initializeGlobal, declareGlobal, debugMsg.
Image - stores information about an image.
filename: the file name relative to IMG_DIR.
data: the image data (UINT8 or double)
Each image structure has to contain the filename or the data
field. It can have both.
type: some text label.
size: the size of the image.
dims: the number of dimensions of the image (2 or 3).
date: time stamp.
See also initializeImage.
Map - 2d data structure with extra information.
origImage: Image from which this map was computed.
label: text label identying the map.
data: 2d array with the map data.
date: time stamp.
parameters: parameters used for generating this map.
See also displayMap, displayMaps.
Pyramid - a multi-resolution pyramid for a particular feature.
origImage: the source image.
label: text label denoting the feature.
type: type of subsampling, one of: 'dyadic','sqrt2','TopDown'.
levels: vector of maps containing the levels of this pyramid.
date: time stamp.
See also makeFeaturePyramids, displayPyramid, runSaliency.
SaliencyParams - set of parameters used for generating a saliency map.
foaSize: size of the focus of attention for disk-IOR.
pyramidType: 'dyadic' or 'sqrt2'.
features: cell array of the features to be used for saliency computation
possible values: 'Color', 'Intensities', 'Orientations', 'Skin','TopDown'.
weights: vector of weights for each feature (same length as features)
IORtype: type of inhibition of return, one of: 'shape','disk','None'.
shapeMode: one of: 'None','shapeSM','shapeCM','shapeFM','shapePyr'.
levelParams: structure with pyramid level parameters.
normtype: Map normalization type, one of: 'None','LocalMax','Iterative'.
numIter: Number of iterations for 'Iterative' normtype.
useRandom: Use random jitter (1) or not (0) for converting coodinates.
segmentComputeType: Method for shape segmentation, one of: 'Fast','LTU'.
IORdecay: decay rate of the inhibitive conductance responsible for inihibiiton
of return per simulation step of 0.1 ms.
smOutputRange: saliency map output in Amperes (1e-9).
noiseAmpl: amplitude of random noise (1e-17).
noiseConst: amplitude of contant noise (1e-14).
gaborParams: structure with parameters for Gabor orientation filters.
oriAngles: vector with orientation angles (in degrees).
oriComputeType: 'efficient' (default) or 'full'
'efficient' only computes orientation filters for the levels that are
actually going to be used in the center-surround computation
'full' computes the orientation filters for all levels
visualizationStyle: style used for visualizing attended locations,
one of: 'Contour', 'ContrastModulate', 'None'.
exclusionMask: struct of type map or 2d mask of numeric or logical type
regions where exclusionMask is ~= 0 are excluded from the saliency
computations by setting all feature maps to 0 in these areas. The
mask is automatically resized to the size of the feature maps.
See also diskIOR, makeGaussianPyramid, makeSaliencyMap, applyIOR, estimateShape,
centerSurround, maxNormalize, winnerToImgCoords, makeGaborFilter,
defaultGaborParams, defaultLevelParams, plotSalientLocation.
levelParams - a structure with parameters for pyramid levels for
center-surround operations
minLevel: lowest pyramid level (starting at 1) for center-surround computations.
maxLevel: highest pyramid level for center-surround.
minDelta: minimum distance (levels) between center and surround.
maxDelta: maximum distance (levels) between center and surround.
mapLevel: pyramid level for all maps, including the saliency map.
See also defaultLevelParams, centerSurround, winnerToImgCoords.
gaborParams - a structure with parameters for Gabor orientation filters.
filterPeriod: the period of the filter in pixels.
elongation: the ratio of length versus width.
filterSize: the size of the filter in pixels.
stddev: the standard deviation of the Gaussian envelope in pixels.
phases: the phase angles to be used.
See also defaultGaborParams, makeGaborFilter, gaborFilterMap, makeOrientationPyramid.
hueParams - describes 2d Gaussian color distribution in CIE space.
muR: mean value in the CR direction.
sigR: standard deviation in the CR direction.
muG: mean value in the CG direction.
sigG: standard deviation in the CG direction.
rho: correlation coefficient between CR and CG.
See also hueDistance, makeHuePyramid, skinHueParams.
saliencyData - a vector of structures for each feature with additional
information from computing the saliency map.
origImage: Image structure of the input image.
label: the feature name.
pyr: a vector of pyramids for this feature.
FM: a vector of feature maps.
csLevels: the center and surround levels used to
compute the feature maps from the pyramids.
CM: the conspicuity map for this feature.
date: time stamp.
See also makeSaliencyMap, estimateShape, runSaliency.
shapeData - information about the shape of the attended regions.
origImage: the Image structure for the source image.
winner: the winning location in saliency map coordinates.
winningMap: the map for the most salient feature at the winner location.
iorMask: the mask used for shape-based inhibition of return.
binaryMap: a binary map of the attended region.
segmentedMap: the winning map segmented by the binary map.
shapeMap: a smoothed version of segmentedMap.
date: time stamp.
See also estimateShape, shapeIOR, applyIOR, plotSalientLocation, runSaliency.
WTA - a winner-take-all neural network.
sm: LIF neuron field for input from the saliency map.
exc: excitatory LIF neurons field.
inhib: inhibitory inter-neuron.
See also initializeWTA, evolveWTA.
LIF - leaky integrate and fire neuron (field).
timeStep: time step for integration (in sec).
Eleak: leak potential (in V).
Eexc: potential for excitatory channels (positive, in V).
Einh: potential for inhibitory channels (negative, in V).
Gleak: leak conductivity (in S).
Gexc: conductivity of excitatory channels (in S).
Ginh: conductivity of inhibitory channels (in S).
GinhDecay: time constant for decay of inhibitory conductivity (in S).
Ginput: input conductivity (in S).
Vthresh: threshold potential for firing (in V).
C: capacity (in F).
time: current time (in sec).
V: current membrane potential (in V) - can be an array for several neurons.
I: current input current (in A) - can be an array for several neurons.
DoesFire: neuron can (1) or cannot (0) fire.
See also defaultLeakyIntFire, evolveLeakyIntFire, initializeWTA.
LTUnetwork - a network of N linear threshold units.
connections: N x N weight matrix, a sparse matrix.
thresholds: 1 x N vector with thresholds for the units.
input_idx: the indices of all input units in the network.
output_idx: the indices of all output units in the network.
numCells: the number of units.
label: a text label fo the network.
See also LTUsimulate, LTUsegmentMap, makeLTUsegmentNetwork.0001 % dataStructures - lists the data structures used in the SaliencyToolbox. 0002 % 0003 % DATA STRUCTURES USED IN THE SALIENCYTOOLBOX 0004 % 0005 % Global variables 0006 % IS_INITIALIZED: flag that initializeGlobal was called. 0007 % IMG_EXTENSIONS: cell arrays with possible extensions for image files. 0008 % DEBUG_FID: file identifier for debugMsg output. 0009 % PD: path delimiter for your operating system. 0010 % BASE_DIR: base directory for data and image locations. 0011 % IMG_DIR: directory for images. 0012 % DATA_DIR: directory for data. 0013 % TMP_DIR: directory for temporary files. 0014 % 0015 % See also initializeGlobal, declareGlobal, debugMsg. 0016 % 0017 % 0018 % Image - stores information about an image. 0019 % filename: the file name relative to IMG_DIR. 0020 % data: the image data (UINT8 or double) 0021 % Each image structure has to contain the filename or the data 0022 % field. It can have both. 0023 % type: some text label. 0024 % size: the size of the image. 0025 % dims: the number of dimensions of the image (2 or 3). 0026 % date: time stamp. 0027 % 0028 % See also initializeImage. 0029 % 0030 % 0031 % Map - 2d data structure with extra information. 0032 % origImage: Image from which this map was computed. 0033 % label: text label identying the map. 0034 % data: 2d array with the map data. 0035 % date: time stamp. 0036 % parameters: parameters used for generating this map. 0037 % 0038 % See also displayMap, displayMaps. 0039 % 0040 % 0041 % Pyramid - a multi-resolution pyramid for a particular feature. 0042 % origImage: the source image. 0043 % label: text label denoting the feature. 0044 % type: type of subsampling, one of: 'dyadic','sqrt2','TopDown'. 0045 % levels: vector of maps containing the levels of this pyramid. 0046 % date: time stamp. 0047 % 0048 % See also makeFeaturePyramids, displayPyramid, runSaliency. 0049 % 0050 % 0051 % SaliencyParams - set of parameters used for generating a saliency map. 0052 % foaSize: size of the focus of attention for disk-IOR. 0053 % pyramidType: 'dyadic' or 'sqrt2'. 0054 % features: cell array of the features to be used for saliency computation 0055 % possible values: 'Color', 'Intensities', 'Orientations', 'Skin','TopDown'. 0056 % weights: vector of weights for each feature (same length as features) 0057 % IORtype: type of inhibition of return, one of: 'shape','disk','None'. 0058 % shapeMode: one of: 'None','shapeSM','shapeCM','shapeFM','shapePyr'. 0059 % levelParams: structure with pyramid level parameters. 0060 % normtype: Map normalization type, one of: 'None','LocalMax','Iterative'. 0061 % numIter: Number of iterations for 'Iterative' normtype. 0062 % useRandom: Use random jitter (1) or not (0) for converting coodinates. 0063 % segmentComputeType: Method for shape segmentation, one of: 'Fast','LTU'. 0064 % IORdecay: decay rate of the inhibitive conductance responsible for inihibiiton 0065 % of return per simulation step of 0.1 ms. 0066 % smOutputRange: saliency map output in Amperes (1e-9). 0067 % noiseAmpl: amplitude of random noise (1e-17). 0068 % noiseConst: amplitude of contant noise (1e-14). 0069 % gaborParams: structure with parameters for Gabor orientation filters. 0070 % oriAngles: vector with orientation angles (in degrees). 0071 % oriComputeType: 'efficient' (default) or 'full' 0072 % 'efficient' only computes orientation filters for the levels that are 0073 % actually going to be used in the center-surround computation 0074 % 'full' computes the orientation filters for all levels 0075 % visualizationStyle: style used for visualizing attended locations, 0076 % one of: 'Contour', 'ContrastModulate', 'None'. 0077 % exclusionMask: struct of type map or 2d mask of numeric or logical type 0078 % regions where exclusionMask is ~= 0 are excluded from the saliency 0079 % computations by setting all feature maps to 0 in these areas. The 0080 % mask is automatically resized to the size of the feature maps. 0081 % 0082 % See also diskIOR, makeGaussianPyramid, makeSaliencyMap, applyIOR, estimateShape, 0083 % centerSurround, maxNormalize, winnerToImgCoords, makeGaborFilter, 0084 % defaultGaborParams, defaultLevelParams, plotSalientLocation. 0085 % 0086 % 0087 % levelParams - a structure with parameters for pyramid levels for 0088 % center-surround operations 0089 % minLevel: lowest pyramid level (starting at 1) for center-surround computations. 0090 % maxLevel: highest pyramid level for center-surround. 0091 % minDelta: minimum distance (levels) between center and surround. 0092 % maxDelta: maximum distance (levels) between center and surround. 0093 % mapLevel: pyramid level for all maps, including the saliency map. 0094 % 0095 % See also defaultLevelParams, centerSurround, winnerToImgCoords. 0096 % 0097 % 0098 % gaborParams - a structure with parameters for Gabor orientation filters. 0099 % filterPeriod: the period of the filter in pixels. 0100 % elongation: the ratio of length versus width. 0101 % filterSize: the size of the filter in pixels. 0102 % stddev: the standard deviation of the Gaussian envelope in pixels. 0103 % phases: the phase angles to be used. 0104 % 0105 % See also defaultGaborParams, makeGaborFilter, gaborFilterMap, makeOrientationPyramid. 0106 % 0107 % 0108 % hueParams - describes 2d Gaussian color distribution in CIE space. 0109 % muR: mean value in the CR direction. 0110 % sigR: standard deviation in the CR direction. 0111 % muG: mean value in the CG direction. 0112 % sigG: standard deviation in the CG direction. 0113 % rho: correlation coefficient between CR and CG. 0114 % 0115 % See also hueDistance, makeHuePyramid, skinHueParams. 0116 % 0117 % 0118 % saliencyData - a vector of structures for each feature with additional 0119 % information from computing the saliency map. 0120 % origImage: Image structure of the input image. 0121 % label: the feature name. 0122 % pyr: a vector of pyramids for this feature. 0123 % FM: a vector of feature maps. 0124 % csLevels: the center and surround levels used to 0125 % compute the feature maps from the pyramids. 0126 % CM: the conspicuity map for this feature. 0127 % date: time stamp. 0128 % 0129 % See also makeSaliencyMap, estimateShape, runSaliency. 0130 % 0131 % 0132 % shapeData - information about the shape of the attended regions. 0133 % origImage: the Image structure for the source image. 0134 % winner: the winning location in saliency map coordinates. 0135 % winningMap: the map for the most salient feature at the winner location. 0136 % iorMask: the mask used for shape-based inhibition of return. 0137 % binaryMap: a binary map of the attended region. 0138 % segmentedMap: the winning map segmented by the binary map. 0139 % shapeMap: a smoothed version of segmentedMap. 0140 % date: time stamp. 0141 % 0142 % See also estimateShape, shapeIOR, applyIOR, plotSalientLocation, runSaliency. 0143 % 0144 % 0145 % WTA - a winner-take-all neural network. 0146 % sm: LIF neuron field for input from the saliency map. 0147 % exc: excitatory LIF neurons field. 0148 % inhib: inhibitory inter-neuron. 0149 % 0150 % See also initializeWTA, evolveWTA. 0151 % 0152 % 0153 % LIF - leaky integrate and fire neuron (field). 0154 % timeStep: time step for integration (in sec). 0155 % Eleak: leak potential (in V). 0156 % Eexc: potential for excitatory channels (positive, in V). 0157 % Einh: potential for inhibitory channels (negative, in V). 0158 % Gleak: leak conductivity (in S). 0159 % Gexc: conductivity of excitatory channels (in S). 0160 % Ginh: conductivity of inhibitory channels (in S). 0161 % GinhDecay: time constant for decay of inhibitory conductivity (in S). 0162 % Ginput: input conductivity (in S). 0163 % Vthresh: threshold potential for firing (in V). 0164 % C: capacity (in F). 0165 % time: current time (in sec). 0166 % V: current membrane potential (in V) - can be an array for several neurons. 0167 % I: current input current (in A) - can be an array for several neurons. 0168 % DoesFire: neuron can (1) or cannot (0) fire. 0169 % 0170 % See also defaultLeakyIntFire, evolveLeakyIntFire, initializeWTA. 0171 % 0172 % 0173 % LTUnetwork - a network of N linear threshold units. 0174 % connections: N x N weight matrix, a sparse matrix. 0175 % thresholds: 1 x N vector with thresholds for the units. 0176 % input_idx: the indices of all input units in the network. 0177 % output_idx: the indices of all output units in the network. 0178 % numCells: the number of units. 0179 % label: a text label fo the network. 0180 % 0181 % See also LTUsimulate, LTUsegmentMap, makeLTUsegmentNetwork. 0182 0183 % This file is part of the SaliencyToolbox - Copyright (C) 2006-2008 0184 % by Dirk B. Walther and the California Institute of Technology. 0185 % See the enclosed LICENSE.TXT document for the license agreement. 0186 % More information about this project is available at: 0187 % http://www.saliencytoolbox.net 0188 0189 more on; 0190 help(mfilename); 0191 more off;