readme

图像中非刚性曲线的蛇形检测算法

This directory contains the test and demonstration programs for the GSNAKE API.

Generally, the programs take the following switches :

	-M	Magnification factor, default 1
	-I	Contour initialisation method by mouse
		[ c - click ] [ d - drag ] [ a - auto ]
	-m	GSNAKE contour mode
		[ o - opened snake] [ c - closed snake ]
	-N	Number of snaxels in a contour
	-n	Number of contours to be localised
	-L	Number of Gaussian pyramid levels to be generated,
		default 2
	-l	Gaussian pyramid level to be accessed
	-S	Stratified search segment spacing
	-H	Numbers of the search segment
	-D	Desired snaxel spacing of an expanded snake
	-E	External energy input type
		[ i - intensity ]
		[ m - edge magnitude ]
		[ e - edge magnitude and direction ]
	-C	Resolution of fine correlation
		[ x - x direction ] [ y - y direction ]
	-r	Rotation, angle in degree
	-t	Translation
		[ x - x direction ] [ y - y direction ]
	-s	Scaling
		[ x - x direction ] [ y - y direction ]
	-d	dilation
		[ x - x direction ] [ y - y direction ]
	-P	Percentage values for histogram specification,
		[ l - low value, default 0.9 ]
		[ h - high value, default 0.95 ]
	-V	Pixel values for histogram specification
		[ l - low value, default 0.2 ]
		[ h - high value, default 0.9 ]
	-e	Exponential value for the transformation function
		of histogram specification, default 1.0
	-T	Image template for correlation
	-U	Starting point of a line or center of circle
		[ x - column coordinate ] [ y - row coordinate ]
	-V	Ending point of a line
		[ x - column coordinate ] [ y - row coordinate ]
	-R	Radius of a circle
		[ x - column coordinate ] [ y - row coordinate ]
	-O	Ouput image/contour file
	-F	Input image/contour file
	
If no switches are given, the default value will be used in the program. The image can be in SUN raster or BIN format. The command line is as below :

	Usage : command	<image>	[option]	or
		command <contour> [option]	or
		command <image> <contour> [option]

<image> refers to the image file, while <contour> refers to the standard GSNAKE contour file. The following gives a brief description of the programs :

Program		Explanation
=======		===========
OOima1	Performs correlation between image and template. The resulting
	image will be shown on an X window and written to an output file.
	Usage :	OOima1 <image> -T<image> [ -M -O<output file> ]
	Example : OOima1 circle.rs -Ttemplate.rs -M2 -Oresult.rs

OOima2	Performs histogram specification of image. The original image and
	resulting image will be shown on an X window.
	Usage : OOima2 <image> [ -M -Pl -Ph -Vl -Vh -E -O<output file> ]
	Example : OOima2 circle.rs -M2 -Oresult.rs

OOima3	Performs cutting and copying of images.
	Usage : OOima3 <image> [ -M -O<output file> ]
	Example : OOima3 circle.rs -Oresult.ras

OOedg1	Computes the edge magnitude and direction of an image. The edge
	magnitude and direction images will be shown on an X window.
	Usage : OOedg0 [ -F<image> -M ]
	Example : OOedg0 -Fcircle.rs -M2

OOpym1	Generates Gaussian pyramid images by default parameters of histogram
	specification. These images include Gaussian blurred images, edge 
	magnitude images and edge direction images.
	Usage : OOpym1 <image> [ -M -L ]
	Example : OOpym1 circle.rs -M2 -L1

OOpym2	Accesses a paticular level of Gaussian pyramid images.
	Usage : OOpym2 <image> [ -M -L -l ]
	Example : OOpym2 akina.rs -L3 -l2

OOsna1	Computes the internal energy of each snaxel. 
	Usage : OOsna1 <contour> [ -M ]
	Example : OOsna1 circle.con -M2

OOsna2	Computes the tangent and normal angle of each snaxel.
	Usage : OOsna2 <contour> [ -M ]
	Example : OOsna2 circle.con -M2

OOcon1	Initializes an opened or a closed contour. The program takes the
	starting and end points of a contour for an opened gsnake, and
	takes the radius for a closed gsnake (circle). The shape matrix
	and internal energy of a contour will also be calculated.
	Usage : OOcon1 [ -m -M -N -Ux -Uy -Vx -Vy -R ]
	Example : OOcon1 -mo -Ux50 -Uy50 -R50

OOcon2	Performs affine transformation of a contour. The internal energy
	before and after transformation is zero. This proves that the shape
	matrix is implicitly invariant to affine transformation.
	Usage : OOcon2 <contour> [ -r -t -s -d -O<output file> ]
	Example : OOcon2 circle.con -r30 -tx20 -sx0.8 -dx1.2  

OOcon3	Initializes an arbitrary contour by mouse or program. The coordinate
	conversion between the image-centered and contour-centered form 
	will also be shown.
	Usage : OOcon3 <image> [ -M -N -I -m ]
	Example : OOcon3 circle.rs -Ic -mo

OOloc1	Provides visualization of the snaxel search locations during energy
	minimization state.
	Usage : OOloc1 <contour> [ -M -S -Hx -Hy ]
	Example : OOloc1 circle.con -S3 -Hx3 -Hy3

OOgho1	Performs localization of a contour on an image by general Hough
	transform (GHT). This is done by first generating Gaussian pyramid
	images. These images are serves as input to GHT. Multiple matched
	contours may also be returned by this program.
	Usage : OOgho1 <image> <contour> [ -M -L -n ]
	Example : OOgho1 akina.rs mouth.con -M2 -L2 -n3

OOgsn1	Performs localization and minimization of a contour on an image.
	Usage : OOgsn1 <image> <contour> [ -M -L -S -D -H -E ]
	Example : OOgsn1 akina mouth.con -L2 -Ee

OOgsn2	Calculates the internal, external and total energy of a gsnake before
	and after manual deformation of the contour by mouse.
	Usage : OOgsn2 <image> [ -M -L -O<output file> ]
	Example : OOgsn2 circle.rs -M2 -L2 -Ocircle.con

OOgsn3	Calculates the internal, external and total energy of a gsnake before
	and after fine correlation around the center of gravity.
	Usage : OOgsn3 <image> [ -M -L -Cx -Cy ]
	Example : OOgsn3 circle.rs -M2 -L2 -Cx3 -Cy3

OOlea1	To learn shape matrix and demonstrate the invariance of a shape matrix.
	For example, use several square images which undergo affine
	transformation to train the shape matrix. The program will then
	regenerate the shape to show that the regenerative shape is still a
	square.
	Usage : OOlea1 <image samples> [ -M -L ]
	Example : OOglea1 rect.rs.* -L2

OOcla1	To classify various templates by marginalization of the distribution,
	MAP probability, match of deformable template, or match of rigid
	template.
	Usage : OOcla1 <image samples>
	Example : OOcla1 rect.rs.1 rect.rs.2