mrpt::vision::CImageConvolution Class Reference

This class performs filtering on a image based on Gaussian or DOG filters. More...

#include <mrpt/vision/CImageConvolution.h>

List of all members.

Static Public Member Functions
static void	convolutionHorzVert (CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CGaussianConvolutionKernel &kernel)
	Performs a 2D convolution of "inImg" using "kernel" and saves the result in "outImg".
static void	convolutionHorizontal (CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CGaussianConvolutionKernel &kernel, TImageROI ROI=TImageROI())
	Performs a horizontal convolution of "inImg" using "kernel" and saves the result in "outImg".
static void	convolutionVertical (CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CGaussianConvolutionKernel &kernel, TImageROI ROI=TImageROI())
	Performs a vertical convolution of "inImg" using "kernel" and saves the result in "outImg".
static void	convolutionSeparate (CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CGaussianConvolutionKernel &kernelHorz, CGaussianConvolutionKernel &kernelVert, TImageROI ROI=TImageROI())
	Performs vertical/horizontal convolution of "inImg" using two different kernels and saves the result in "outImg".
static void	convolutionGeneric (const CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CMatrix &kernel, unsigned int x_min=0, unsigned int x_length=0, unsigned int y_min=0, unsigned int y_length=0)
	Performs convolution of "inImg" using a generic kernel and saves the result in "outImg".
static void	convolutionGeneric (const CMRPTImageFloat &inImg, CMRPTImageFloat &outImg, CMatrix &kernel, TImageROI ROI=TImageROI())
static void	computeGradients (CMRPTImageFloat &inImg, float sigma, CMRPTImageFloat &outGradientX, CMRPTImageFloat &outGradientY)
	Computes the gradient of an image for a given sigma (Gaussian + DOG filtering).
static void	computeSmoothedImage (CMRPTImageFloat &inImg, float sigma, CMRPTImageFloat &outImg, TImageROI ROI=TImageROI())
	Computes the smoothed version of an image for a given sigma (Gaussian filtering).

---

Detailed Description

This class performs filtering on a image based on Gaussian or DOG filters.

See also:

vision::CImageConvolution, vision

Definition at line 45 of file CImageConvolution.h.

---

Member Function Documentation

static void mrpt::vision::CImageConvolution::computeGradients	(	CMRPTImageFloat &	inImg,
		float	sigma,
		CMRPTImageFloat &	outGradientX,
		CMRPTImageFloat &	outGradientY
	)			[static]

Computes the gradient of an image for a given sigma (Gaussian + DOG filtering).

NOTES:

• "inImg" and "outImg" MUST BE different objects!
• Internally, the kernels are saved between calls, thus they are computed just the first time for each sigma value.

static void mrpt::vision::CImageConvolution::computeSmoothedImage	(	CMRPTImageFloat &	inImg,
		float	sigma,
		CMRPTImageFloat &	outImg,
		TImageROI	ROI = TImageROI()
	)			[static]

Computes the smoothed version of an image for a given sigma (Gaussian filtering).

NOTES:

• "inImg" and "outImg" MUST BE different objects!
• Internally, the kernel is saved between calls, thus it is computed just the first time for each sigma value.

static void mrpt::vision::CImageConvolution::convolutionGeneric	(	const CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CMatrix &	kernel,
		TImageROI	ROI = TImageROI()
	)			[static]

static void mrpt::vision::CImageConvolution::convolutionGeneric	(	const CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CMatrix &	kernel,
		unsigned int	x_min = 0,
		unsigned int	x_length = 0,
		unsigned int	y_min = 0,
		unsigned int	y_length = 0
	)			[static]

Performs convolution of "inImg" using a generic kernel and saves the result in "outImg".

ROI = [x_min, x_min+x_length-1] x [y_min, y_min+y_length-1] NOTE: "inImg" and "outImg" MUST BE different objects! by FAMD @ JAN-2007

static void mrpt::vision::CImageConvolution::convolutionHorizontal	(	CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CGaussianConvolutionKernel &	kernel,
		TImageROI	ROI = TImageROI()
	)			[static]

Performs a horizontal convolution of "inImg" using "kernel" and saves the result in "outImg".

NOTE: "inImg" and "outImg" MUST BE different objects!

static void mrpt::vision::CImageConvolution::convolutionHorzVert	(	CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CGaussianConvolutionKernel &	kernel
	)			[static]

Performs a 2D convolution of "inImg" using "kernel" and saves the result in "outImg".

NOTES:

• "inImg" and "outImg" MUST BE different objects!
• The convolution is performed in two separable steps: vertical and horizontal.

static void mrpt::vision::CImageConvolution::convolutionSeparate	(	CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CGaussianConvolutionKernel &	kernelHorz,
		CGaussianConvolutionKernel &	kernelVert,
		TImageROI	ROI = TImageROI()
	)			[static]

Performs vertical/horizontal convolution of "inImg" using two different kernels and saves the result in "outImg".

NOTE: "inImg" and "outImg" MUST BE different objects!

static void mrpt::vision::CImageConvolution::convolutionVertical	(	CMRPTImageFloat &	inImg,
		CMRPTImageFloat &	outImg,
		CGaussianConvolutionKernel &	kernel,
		TImageROI	ROI = TImageROI()
	)			[static]

Performs a vertical convolution of "inImg" using "kernel" and saves the result in "outImg".

NOTE: "inImg" and "outImg" MUST BE different objects!

Todo:

JL: There is bug in "convolutionVertical" (& probably convolutionHorizontal) for images much smaller than the kernel (run MonoSLAM!)