This is a color definition class. It requires training and supports both normalized RGB and HSI format. 2D histograms of the color distribution in the given format are built based on the training data. Can be used to define any (and even multiple) color.


Public Member Functions
	tlColorModel ()
	Constructor.
	tlColorModel (char *filebase)
virtual	~tlColorModel ()
	Destructor.
int	reset ()
	Reset color model.
int	finalize ()
int	save (char *filebase)
int	load (char *filebase)
int	train (tlImage *image)
int	train (tlImage image, tlPoint point)
int	train (tlImage image, tlChain chain)
int	filter (tlImage *image, int threshold=TL_COLOR_MODEL_DETECTION_THRESHOLD)
int	filter (tlImage image, tlRect rect, int threshold=TL_COLOR_MODEL_DETECTION_THRESHOLD)
int	test (tlColor *color, int threshold=TL_COLOR_MODEL_DETECTION_THRESHOLD)

Constructor & Destructor Documentation

tlColorModel::tlColorModel ( char * filebase )

Constructor.

Parameters:

filebase

basename for color model file to upload.

Member Function Documentation

int tlColorModel::finalize ( )

Once training is finished, call this method to scale the histogram correctly. The threshold argument for the filter() method will then be expressed in percent, as opposed to pixel count.

int tlColorModel::save ( char * filebase )

Save histograms to disk.

Parameters:

filebase

base name for the hitogram files.

int tlColorModel::load ( char * filebase )

Load histograms from disk.

Parameters:

filebase

base name for the hitogram files.

int tlColorModel::train ( tlImage * image )

Update the histograms based on a segmented image of either format.

Parameters:

image

segmented image, only pixels of the desired color are non-zero.

Note:: Only the histogram of the image format is updated. User should call finalize() once training is finished.

int tlColorModel::train	(	tlImage *	image,
		tlPoint *	point
	)

Add the contribution of a single pixel from a training image to the color histogram.

Parameters:

	image	the image containing the pixel of interest.
	point	the location of the pixel to train on.

Note:: Only the histogram of the image format is updated. A better (and faster) way to build the color distribution is to use a pre-segmented image with train(tlImage *image). User should call finalize() once training is finished.

int tlColorModel::train	(	tlImage *	image,
		tlChain *	chain
	)

Add the contribution of a number of pixels from a training image to the color histogram.

Parameters:

	image	the image containing the pixel of interest.
	chain	the locations of the pixels to train on.

Note:: Only the histogram of the image format is updated. A better (and faster) way to build the color distribution is to use a pre-segmented image with train(tlImage *image). User should call finalize() once training is finished.

int tlColorModel::filter	(	tlImage *	image,
		int	threshold = `TL_COLOR_MODEL_DETECTION_THRESHOLD`
	)

Filter a color image using the (thresholded) color distribution contained in the histograms. That is, only pixels with color value that contribute for more than 'threshold' percent to the histogram will be kept.

Parameters:

	image	the image to be filtered.
	threshold	percentage or minimal pixel count.

Note:: Threshold represents a percentage only if finalize() is called after the training is over. Otherwise, threshold must be expressed in pixel counts.

int tlColorModel::filter	(	tlImage *	image,
		tlRect *	rect,
		int	threshold = `TL_COLOR_MODEL_DETECTION_THRESHOLD`
	)

Filter a portion of a color image using the (thresholded) color distribution contained in the histograms. That is, only pixels with color value that contribute for more than 'threshold' percent to the histogram will be kept.

Parameters:

	image	the image to be filtered.
	rect	portion of the image to process
	threshold	percentage or minimal pixel count.

Note:: Threshold represents a percentage only if finalize() is called after the training is over. Otherwise, threshold must be expressed in pixel counts.

int tlColorModel::test	(	tlColor *	color,
		int	threshold = `TL_COLOR_MODEL_DETECTION_THRESHOLD`
	)

Test whether a given color accounts for more than a given threshold in the color histograms.

Parameters:

	color	the color to test for
	threshold	the histogram threshold value

tlColorModel Class Reference

Detailed Description

Public Member Functions

Constructor & Destructor Documentation

Member Function Documentation