Classes
class	kjb::Color_histogram
	Class to compute an RGB colour histogram over an image or a rectangular portion of it. The histogram is normalized. We use the same number of bins for each of the channels (r,g,b). It is easy to extend this class so that it computes such histogram over a segment of a shape other than rectangular. More...

class	kjb::Colormap

class	kjb::Filter
	Filter class. More...

struct	kjb::PixelHSVA
	Alternative Pixel using hue, saturation, value, and opacity (alpha). More...

class	kjb::Image
	Wrapped version of the C struct KJB_image. More...

class	kjb::Fftw_image_convolution
	this is a simple adaptation of Fftw_convolution_2d to Image input. More...

class	kjb::Opengl_framebuffer_image

struct	kjb::PixelRGBA
	Wrapped version of the C struct Pixel, with Alpha (opacity). More...

Typedefs
typedef boost::shared_ptr< Image >	kjb::Image_ptr

Functions
Image	kjb::make_collage (Image const const input_images, int num_horizontal, int num_vertical)
	wrapper for kjb_c::make_image_collage More...

Image	kjb::make_collage (Image const *input_img_array, int num_horizontal, int num_vertical)
	also works for an array of Image objects More...

void	kjb::write_histograms (const std::vector< Color_histogram > &chs, const std::string &file_name)

void	kjb::read_histograms (std::vector< Color_histogram > &chs, const std::string &file_name)

Filter	kjb::gaussian_filter (double sigma, int size)
	Create a Gaussian filter with given sigma and size. More...

Filter	kjb::gaussian_filter (double sigma)
	Create a Gaussian filter with given sigma. More...

Filter	kjb::laplacian_of_gaussian_filter (int size, double sigma)
	Create a Laplacian of Gaussian filter. More...

Image	kjb::operator* (const Image &image, const Filter &kernel)
	Convolve an image with a filter. More...

Image	kjb::gauss_sample_image (const Image &in, int resolution, double sigma)
	this wraps C function kjb_c::gauss_sample_image (q.v.). More...

Vector	kjb::hsluma_space (const kjb_c::Pixel &p)
	Compute the spatial coordinates of a pixel in HSY color space. More...

kjb_c::Pixel	kjb::get_pixel_from_hsluma_space (const Vector &hsy)
	This clampfully transforms a point in HSY space to RGB. More...

int	kjb::get_pixel_from_hsluma_space (const Vector &hsy, kjb_c::Pixel *p)
	This might transform a point in HSY space to RGB. More...

Image	kjb::scale_image (const Image &i, double factor)
	Scale image size by factor, i.e., enlarge or shrink. More...

Image	kjb::get_inverted (const Image &)

Image	kjb::rgb_matrices_to_image (const Matrix &red_channel, const Matrix &green_channel, const Matrix &blue_channel)
	Contruct an image from three matrices representing red, green, blue channels. More...

Matrix	kjb::to_grayscale_matrix (const Image &i)

Image	kjb::operator* (const Image &op1, double op2)
	Scale an image in channel space, yielding a new image. More...

Image	kjb::operator/ (const Image &op1, double op2)
	Scale an image in channel space, yielding a new image. More...

Image	kjb::operator+ (const Image &op1, const Image &op2)
	Add two images. More...

Image	kjb::operator- (const Image &im1, const Image &im2)
	Subtract two images. More...

void	kjb::enable_transparency (Image &i)
	set Image flag, to indicate that the 'alpha' channel is meaningful. More...

void	kjb::disable_transparency (Image &i)
	clear Image flag, to indicate that the 'alpha' channel is not meaningful. More...

bool	kjb::is_transparency_enabled (const Image &i)
	test the Image flag, returning true if the 'alpha' channel is meaningful. More...

Image	kjb::matrix_to_max_contrast_8bit_bw_image (const kjb::Matrix &m)
	convert a matrix to a black-and-white image, with enhanced contrast. More...

Image	kjb::matrix_to_color_image (const kjb::Matrix &m, const kjb::Colormap &map="jet")
	Converts a matrix to a high-contrast color image, mapping values to colors using a kjb::Colormap. Equivalent to Matlab's imagesc() function. More...

Image	kjb::imagesc (const kjb::Matrix &data, const kjb::Colormap &map="jet")
	Ported version of Matlab's "imagesc()" function; alias of matrix_to_color_image. More...

std::string	kjb::pixel_as_hex_triplet_string (const kjb_c::Pixel *p)
	express color as HTML-style hex triplet, e.g., "#FFCC00," of pointer More...

PixelRGBA	kjb::operator* (const PixelRGBA &p, const PixelRGBA &q)
	Multiply two pixels together. More...

PixelRGBA	kjb::operator* (const PixelRGBA &p, double k)
	Scale a pixel by a floating point value on the right. More...

PixelRGBA	kjb::operator* (double k, const PixelRGBA &p)
	Scale a pixel by a floating point value on the left. More...

kjb_c::Pixel	kjb::abs (const kjb_c::Pixel &p)
	Take the channel-wise absolute value of a kjb_c::Pixel. More...

std::string	kjb::pixel_as_hex_triplet_string (const kjb_c::Pixel &p)
	express color as HTML-style hex triplet, e.g., "#FFCC00," reference More...

template<class InputIterator >
Vector	kjb::Image::intensity_histogram (InputIterator first, InputIterator last) const
	Computes the intensity histogram of a region – given by a range of pairs – of this image. More...

Detailed Description

This group include the Image class and all closely-related classes and functions pertaining to low-level operations on and manipulation of images.

Typedef Documentation

typedef boost::shared_ptr<Image> kjb::Image_ptr

Convenience typedef for boost::shared_ptr<Image>.

Function Documentation

kjb_c::Pixel kjb::abs ( const kjb_c::Pixel & p )

inline

Take the channel-wise absolute value of a kjb_c::Pixel.

void kjb::disable_transparency ( Image & i )

inline

clear Image flag, to indicate that the 'alpha' channel is not meaningful.

void kjb::enable_transparency ( Image & i )

inline

set Image flag, to indicate that the 'alpha' channel is meaningful.

Image kjb::gauss_sample_image	(	const Image &	in,
		int	resolution,
		double	sigma
	)

this wraps C function kjb_c::gauss_sample_image (q.v.).

Filter kjb::gaussian_filter	(	double	sigma,
		int	size
	)

Create a Gaussian filter with given sigma and size.

Filter kjb::gaussian_filter ( double sigma )

inline

Create a Gaussian filter with given sigma.

Image kjb::get_inverted ( const Image & i )

inline

kjb_c::Pixel kjb::get_pixel_from_hsluma_space ( const Vector & hsy )

This clampfully transforms a point in HSY space to RGB.

Returns: A pixel derived from input HSY coordinates, possibly clamped.

Please see the comments for get_pixel_from_hsluma_space( const kjb::Vector&, kjb_c::Pixel* ).

If the hue-saturation-luma coordinates are outside the HSY polyhedron, this still returns a pixel, but the resulting R,G,B values are clamped to the range 0 to 255. Alpha (opacity) is set to the default value used by class PixelRGBA.

int kjb::get_pixel_from_hsluma_space	(	const Vector &	hsy,
		kjb_c::Pixel *	p
	)

This might transform a point in HSY space to RGB.

Returns: EXIT_SUCCESS iff the values in hsy are valid coordinates inside the HSY (Y=luma) color polyhedron. Otherwise, EXIT_FAILURE.

Input hsy is a vector of cartesian coordinates where arctan( hsy[1] / hsy[0] ) is hue, sqrt(hsy[0]**2 + hsy[1]**2) is saturation, and hsy[2] is "luma." (Luma is a tweak to brightness to compensate for the differences in perceived brightness betwen red, green, and blue.) Valid HSY values are necessarily inside a cuboid with opposite corners (-1,-1,0) and (1,1,1); but being in this range is not sufficient. The HSY color space is a polyhedron inside that cuboid.

Iff input hsy is outside the polyhedron, the return value is EXIT_FAILURE.

If the input vector is not within the HSY color polyhedron, then pointer p is not touched. Otherwise, and if p is not equal to null, the resulting pixel is stored in *p. If you do not need the pixel, it is safe to pass in a null value, and this function will act like a predicate.

Vector kjb::hsluma_space ( const kjb_c::Pixel & p )

Compute the spatial coordinates of a pixel in HSY color space.

The space I am using here is described in the Wikipedia article linked above. Please read about the HSL bicone: a bicone shape is like a top generated by revolving a diamond. This forms a 3D shape: picture yourself holding an empty ice cream cone, and on top of that you place an inverted empty ice cream cone. Lightness L runs along the central axis of rotation: the pointy bottom is black, the top tip is white. Hue H runs around the circumference, basically in a cyclical, rainbow order; 0 degrees is for reddish colors, 180 degrees is for cyanic colors. Saturation describes a point's distance from the central axis; at a radius of 0 the colors are only black, white, and grays. Farther from the axis the colors grow more vivid.

Except that I want to use NTSC 601 for lightness, instead of the more conventional I=(max{r,g,b}+min{r,g,b})/2. NTSC 601 defines "luma" Y as an uneven mix of red, green, and blue, in recognition of the fact that perceptually, green seems brighter than red, and both seem dimmer than blue, for typical computer shades of red, green, and blue. This distorts the shape from a bicone. The space remains convex.

In any case, this function returns the cartesian coordinates (not cylindrical!) of color points within this space, in the form of a vector v. If you think of hue H as an angle from 0 to 359, then H is basically arctan(v[1] / v[0]). v[2] is luma; zero luma is black, unity luma is white. Components v[0], v[1] are in the closed interval [-1,1]. Component v[2] is in closed interval [0,1]. Not all points in the unit cube of that space are reachable from the RGB cube!

This is basically an invertible matrix transformation: vector (r,b,g)' times matrix [ [1 -.5 -.5]; [0 .866 -.866]; [.3 .59 .11] ] <===(octave notation). The inverse of this matrix is hard-coded into (overloaded) functions PixelRGBA_from_HSLuma_space().

Image kjb::imagesc	(	const kjb::Matrix &	data,
		const kjb::Colormap &	map = `"jet"`
	)

inline

Ported version of Matlab's "imagesc()" function; alias of matrix_to_color_image.

template<class InputIterator >

Vector kjb::Image::intensity_histogram	(	InputIterator	first,
		InputIterator	last
	)		const

Computes the intensity histogram of a region – given by a range of pairs – of this image.

Template Parameters

InputIterator An iterator whose value_type is a pair.

bool kjb::is_transparency_enabled ( const Image & i )

inline

test the Image flag, returning true if the 'alpha' channel is meaningful.

Filter kjb::laplacian_of_gaussian_filter	(	int	size,
		double	sigma
	)

Create a Laplacian of Gaussian filter.

Image kjb::make_collage	(	Image const const	input_images,
		int	num_horizontal,
		int	num_vertical
	)

wrapper for kjb_c::make_image_collage

Image kjb::make_collage	(	Image const *	input_img_array,
		int	num_horizontal,
		int	num_vertical
	)

also works for an array of Image objects

Image kjb::matrix_to_color_image	(	const kjb::Matrix &	m,
		const kjb::Colormap &	map
	)

Converts a matrix to a high-contrast color image, mapping values to colors using a kjb::Colormap. Equivalent to Matlab's imagesc() function.

Maps matrix values to [0,1], and transforms them to colors using the provided colormap. The following strings may be passed to specify preset colormaps, which are identical to those in Matlab:

Using "gray" should give similar results to matrix_to_max_contast_8bit_bw_image()

Custom colormaps may also by used by passing a kjb::Colormap object instead of as string.

Image kjb::matrix_to_max_contrast_8bit_bw_image ( const kjb::Matrix & m )

convert a matrix to a black-and-white image, with enhanced contrast.

Parameters

m	input matrix to visualize

Returns: Image with all pixels adjusted for enhanced contrast

This wraps the C function kjb_c::matrix_to_max_contrast_8bit_bw_image().

Image kjb::operator*	(	const Image &	image,
		const Filter &	kernel
	)

Convolve an image with a filter.

PixelRGBA kjb::operator*	(	const PixelRGBA &	p,
		const PixelRGBA &	q
	)

inline

Multiply two pixels together.

This simply performs an channel-wise product.

Warning: Don't forget to clamp() if you think you might need to.

PixelRGBA kjb::operator*	(	const PixelRGBA &	p,
		double	k
	)

inline

Scale a pixel by a floating point value on the right.

Warning: Don't forget to clamp() if you think you might need to.

PixelRGBA kjb::operator*	(	double	k,
		const PixelRGBA &	p
	)

inline

Scale a pixel by a floating point value on the left.

Warning: Don't forget to clamp() if you think you might need to.

Image kjb::operator*	(	const Image &	op1,
		double	op2
	)

inline

Scale an image in channel space, yielding a new image.

Image kjb::operator+	(	const Image &	op1,
		const Image &	op2
	)

inline

Add two images.

Image kjb::operator-	(	const Image &	im1,
		const Image &	im2
	)

inline

Subtract two images.

Image kjb::operator/	(	const Image &	op1,
		double	op2
	)

inline

Scale an image in channel space, yielding a new image.

Exceptions

Divide_by_zero if the given value is zero.

std::string kjb::pixel_as_hex_triplet_string ( const kjb_c::Pixel * p )

inline

express color as HTML-style hex triplet, e.g., "#FFCC00," of pointer

std::string kjb::pixel_as_hex_triplet_string ( const kjb_c::Pixel & p )

inline

express color as HTML-style hex triplet, e.g., "#FFCC00," reference

void kjb::read_histograms	(	std::vector< Color_histogram > &	chs,
		const std::string &	file_name
	)

Image kjb::rgb_matrices_to_image	(	const Matrix &	red_channel,
		const Matrix &	green_channel,
		const Matrix &	blue_channel
	)

Contruct an image from three matrices representing red, green, blue channels.

Exceptions

KJB_error,possibly,if matrices not the same size (see below).

This just wraps a call to kjb_c::rgb_matrices_to_image(). If the inputs are not all identically sized, the bug handler is called. If the bug handler returns, a KJB_error is thrown.

Note that if you want your Image to behave like with 8-bit channel values, then you are obliged to scale the input matrices to the range 0.0 to 255.0.

Image kjb::scale_image	(	const Image &	i,
		double	factor
	)

Scale image size by factor, i.e., enlarge or shrink.

Parameters

i	input image to scale
factor	the enlargement factor to apply, e.g., 2 means to double the width and height. 0.5 means to halve the width and height.

Returns: The scaled image.

Exceptions

KJB_error if the scaling fails

If the enlargement factor is in the range 0.9999 to 1.0001, this function just returns a copy of the input, without any scaling.

Matrix kjb::to_grayscale_matrix ( const Image & i )

void kjb::write_histograms	(	const std::vector< Color_histogram > &	chs,
		const std::string &	file_name
	)

Classes

Typedefs

Functions

Detailed Description

Typedef Documentation

Function Documentation