Integral Image

Collaboration diagram for Integral Image:

Functions
template<class V1 , class V2 , class T , class M >
void	cuv::integral_img::integral_image (cuv::tensor< V1, T, M > &dst, const cuv::tensor< V2, T, M > &src)
	calculate the integral image
template<class V1 , class V2 , class T , class M >
void	cuv::integral_img::scan (cuv::tensor< V1, T, M > &dst, const cuv::tensor< V2, T, M > &src)
	integrate rows of an image
template<class V , class M >
void	cuv::integral_img::integral_image_4d (cuv::tensor< V, M > &dst, const cuv::tensor< V, M > &src)
	calculates many integral images in parallel, for data given in format required by Alex' convolutions.

Detailed Description

Function Documentation

template<class V1 , class V2 , class T , class M >

void cuv::integral_img::integral_image	(	cuv::tensor< V1, T, M > &	dst,
		const cuv::tensor< V2, T, M > &	src
	)

calculate the integral image

this applies

See Also

scan twice, transposing in between.

Parameters

src	source
dst	destination

template<class V , class M >

void cuv::integral_img::integral_image_4d	(	cuv::tensor< V, M > &	dst,
		const cuv::tensor< V, M > &	src
	)

calculates many integral images in parallel, for data given in format required by Alex' convolutions.

The input (and output) is assumed to be row-major and

nChannels x nRows x nCols x nImages.

every channel of every image is integrated separately.

We compute the /exclusive/ scan, s.t. the output is

nChannels x (nRows+1) x (nCols+1) x nImages.

Parameters

src	source
dst	destination

template<class V1 , class V2 , class T , class M >

void cuv::integral_img::scan	(	cuv::tensor< V1, T, M > &	dst,
		const cuv::tensor< V2, T, M > &	src
	)

integrate rows of an image

Parameters

src	source
dst	destination