boofcv.alg.geo

Class PerspectiveOps

java.lang.Object

boofcv.alg.geo.PerspectiveOps

public class PerspectiveOps
extends Object

Functions related to perspective geometry and intrinsic camera calibration.

Author:

Peter Abeles

Constructor Summary

Constructors

Constructor and Description
PerspectiveOps()

Method Summary

Methods

Modifier and Type	Method and Description
static PointTransform_F32	adjustIntrinsic_F32(PointTransform_F32 distortPixel, boolean forwardTran, IntrinsicParameters parameters, org.ejml.data.DenseMatrix64F adjustMatrix, IntrinsicParameters adjustedParam) Creates a new PointTransform_F32 which is the same as applying a homography transform and another arbitrary transform.
static org.ejml.data.DenseMatrix64F	calibrationMatrix(double fx, double fy, double skew, double xc, double yc) Given the intrinsic parameters create a calibration matrix
static org.ejml.data.DenseMatrix64F	calibrationMatrix(IntrinsicParameters param, org.ejml.data.DenseMatrix64F K) Given the intrinsic parameters create a calibration matrix
static georegression.struct.point.Point2D_F64	convertNormToPixel(org.ejml.data.DenseMatrix64F K, georegression.struct.point.Point2D_F64 norm, georegression.struct.point.Point2D_F64 pixel) Convenient function for converting from normalized image coordinates to the original image pixel coordinate.
static georegression.struct.point.Point2D_F64	convertNormToPixel(IntrinsicParameters param, double x, double y, georegression.struct.point.Point2D_F64 pixel) Convenient function for converting from normalized image coordinates to the original image pixel coordinate.
static georegression.struct.point.Point2D_F64	convertNormToPixel(IntrinsicParameters param, georegression.struct.point.Point2D_F64 norm, georegression.struct.point.Point2D_F64 pixel) Convenient function for converting from normalized image coordinates to the original image pixel coordinate.
static georegression.struct.point.Point2D_F64	convertPixelToNorm(org.ejml.data.DenseMatrix64F K, georegression.struct.point.Point2D_F64 pixel, georegression.struct.point.Point2D_F64 norm) Convenient function for converting from original image pixel coordinate to normalized< image coordinates.
static georegression.struct.point.Point2D_F64	convertPixelToNorm(IntrinsicParameters param, georegression.struct.point.Point2D_F64 pixel, georegression.struct.point.Point2D_F64 norm) Convenient function for converting from original image pixel coordinate to normalized< image coordinates.
static org.ejml.data.DenseMatrix64F	createCameraMatrix(org.ejml.data.DenseMatrix64F R, georegression.struct.point.Vector3D_F64 T, org.ejml.data.DenseMatrix64F K, org.ejml.data.DenseMatrix64F ret) Create a 3x4 camera matrix.
static IntrinsicParameters	matrixToParam(org.ejml.data.DenseMatrix64F K, int width, int height, boolean flipY, IntrinsicParameters param) Converts a calibration matrix into intrinsic parameters
static georegression.struct.point.Point2D_F64	renderPixel(org.ejml.data.DenseMatrix64F worldToCamera, georegression.struct.point.Point3D_F64 X) Computes the image coordinate of a point given its 3D location and the camera matrix.
static georegression.struct.point.Point2D_F64	renderPixel(georegression.struct.se.Se3_F64 worldToCamera, org.ejml.data.DenseMatrix64F K, georegression.struct.point.Point3D_F64 X) Renders a point in world coordinates into the image plane in pixels or normalized image coordinates.
static void	scaleIntrinsic(IntrinsicParameters param, double scale) Multiplies each element of the intrinsic parameters by the provided scale factor.
static void	splitAssociated(List<AssociatedPair> pairs, List<georegression.struct.point.Point2D_F64> view1, List<georegression.struct.point.Point2D_F64> view2) Takes a list of AssociatedPair as input and breaks it up into two lists for each view.
static void	splitAssociated(List<AssociatedTriple> pairs, List<georegression.struct.point.Point2D_F64> view1, List<georegression.struct.point.Point2D_F64> view2, List<georegression.struct.point.Point2D_F64> view3) Takes a list of AssociatedTriple as input and breaks it up into three lists for each view.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PerspectiveOps

public PerspectiveOps()

Method Detail

scaleIntrinsic

public static void scaleIntrinsic(IntrinsicParameters param,
                  double scale)

Multiplies each element of the intrinsic parameters by the provided scale factor. Useful if the image has been rescaled.

Parameters:

param - Intrinsic parameters

scale - Scale factor that input image is being scaled by.

adjustIntrinsic_F32

public static PointTransform_F32 adjustIntrinsic_F32(PointTransform_F32 distortPixel,
                                     boolean forwardTran,
                                     IntrinsicParameters parameters,
                                     org.ejml.data.DenseMatrix64F adjustMatrix,
                                     IntrinsicParameters adjustedParam)

Creates a new PointTransform_F32 which is the same as applying a homography transform and another arbitrary transform. A typical application is removing lens distortion. The order that each transform is applied depends on if the arbitrary transform is forward or reverse transform. A new set of camera parameters is computed to account for the adjustment.

When removing camera distortion, the undistorted image is likely to have a different shape not entirely enclosed by the original image. This can be compensated for by transforming the undistorted image using a homography transform. Typically this will translate and scale the undistorted image. The end result is a new virtual camera which has the adjusted intrinsic camera parameters.

The returned transform:
λx = A*K*[R|T]X
where A is the homography.

Parameters:

distortPixel - Transform that distorts the pixels in an image.

forwardTran - If true then the distortion expects undistorted pixels as input, false means it expects distorted pixels as input.

parameters - Original intrinsic camera parameters

adjustMatrix - Invertible homography

adjustedParam - The new intrinsic calibration matrix.

Returns:

The new transform.

calibrationMatrix

public static org.ejml.data.DenseMatrix64F calibrationMatrix(double fx,
                                             double fy,
                                             double skew,
                                             double xc,
                                             double yc)

Given the intrinsic parameters create a calibration matrix

Parameters:

fx - Focal length x-axis in pixels

fy - Focal length y-axis in pixels

skew - skew in pixels

xc - camera center x-axis in pixels

yc - center center y-axis in pixels

Returns:

Calibration matrix 3x3

calibrationMatrix

public static org.ejml.data.DenseMatrix64F calibrationMatrix(IntrinsicParameters param,
                                             org.ejml.data.DenseMatrix64F K)

Given the intrinsic parameters create a calibration matrix

Parameters:

param - Intrinsic parameters structure that is to be converted into a matrix

K - Storage for calibration matrix, must be 3x3. If null then a new matrix is declared

Returns:

Calibration matrix 3x3

matrixToParam

public static IntrinsicParameters matrixToParam(org.ejml.data.DenseMatrix64F K,
                                int width,
                                int height,
                                boolean flipY,
                                IntrinsicParameters param)

Converts a calibration matrix into intrinsic parameters

Parameters:

K - Camera calibration matrix.

width - Image width in pixels

height - Image height in pixels

flipY - When calibrated was the y-axis adjusted with: y = (height - y - 1)

param - Where the intrinsic parameter are written to. If null then a new instance is declared.

Returns:

IntrinsicParameters structure.

convertNormToPixel

public static georegression.struct.point.Point2D_F64 convertNormToPixel(IntrinsicParameters param,
                                                        double x,
                                                        double y,
                                                        georegression.struct.point.Point2D_F64 pixel)

Convenient function for converting from normalized image coordinates to the original image pixel coordinate. If speed is a concern then NormalizedToPixel_F64 should be used instead.

NOTE: Lens distortion handled!

Parameters:

param - Intrinsic camera parameters

x - X-coordinate of normalized.

y - Y-coordinate of normalized.

pixel - Optional storage for output. If null a new instance will be declared.

Returns:

pixel image coordinate

convertNormToPixel

public static georegression.struct.point.Point2D_F64 convertNormToPixel(IntrinsicParameters param,
                                                        georegression.struct.point.Point2D_F64 norm,
                                                        georegression.struct.point.Point2D_F64 pixel)

Convenient function for converting from normalized image coordinates to the original image pixel coordinate. If speed is a concern then NormalizedToPixel_F64 should be used instead.

NOTE: Lens distortion handled!

NOTE: norm and pixel can be the same instance.

Parameters:

param - Intrinsic camera parameters

norm - Normalized image coordinate.

pixel - Optional storage for output. If null a new instance will be declared.

Returns:

pixel image coordinate

convertNormToPixel

public static georegression.struct.point.Point2D_F64 convertNormToPixel(org.ejml.data.DenseMatrix64F K,
                                                        georegression.struct.point.Point2D_F64 norm,
                                                        georegression.struct.point.Point2D_F64 pixel)

Convenient function for converting from normalized image coordinates to the original image pixel coordinate. If speed is a concern then NormalizedToPixel_F64 should be used instead.

NOTE: Lens distortion handled!

NOTE: norm and pixel can be the same instance.

Parameters:

K - Intrinsic camera calibration matrix

norm - Normalized image coordinate.

pixel - Optional storage for output. If null a new instance will be declared.

Returns:

pixel image coordinate

convertPixelToNorm

public static georegression.struct.point.Point2D_F64 convertPixelToNorm(IntrinsicParameters param,
                                                        georegression.struct.point.Point2D_F64 pixel,
                                                        georegression.struct.point.Point2D_F64 norm)

Convenient function for converting from original image pixel coordinate to normalized< image coordinates. If speed is a concern then PixelToNormalized_F64 should be used instead.

NOTE: Lens distortion is not removed!

NOTE: norm and pixel can be the same instance.

Parameters:

param - Intrinsic camera parameters

pixel - Pixel coordinate

norm - Optional storage for output. If null a new instance will be declared.

Returns:

normalized image coordinate

convertPixelToNorm

public static georegression.struct.point.Point2D_F64 convertPixelToNorm(org.ejml.data.DenseMatrix64F K,
                                                        georegression.struct.point.Point2D_F64 pixel,
                                                        georegression.struct.point.Point2D_F64 norm)

Convenient function for converting from original image pixel coordinate to normalized< image coordinates. If speed is a concern then PixelToNormalized_F64 should be used instead.

NOTE: norm and pixel can be the same instance.

Parameters:

K - Intrinsic camera calibration matrix

pixel - Pixel coordinate.

norm - Optional storage for output. If null a new instance will be declared.

Returns:

normalized image coordinate

renderPixel

public static georegression.struct.point.Point2D_F64 renderPixel(georegression.struct.se.Se3_F64 worldToCamera,
                                                 org.ejml.data.DenseMatrix64F K,
                                                 georegression.struct.point.Point3D_F64 X)

Renders a point in world coordinates into the image plane in pixels or normalized image coordinates.

Parameters:

worldToCamera - Transform from world to camera frame

K - Optional. Intrinsic camera calibration matrix. If null then normalized image coordinates are returned.

X - 3D Point in world reference frame..

Returns:

2D Render point on image plane or null if it's behind the camera

renderPixel

public static georegression.struct.point.Point2D_F64 renderPixel(org.ejml.data.DenseMatrix64F worldToCamera,
                                                 georegression.struct.point.Point3D_F64 X)

Computes the image coordinate of a point given its 3D location and the camera matrix.

Parameters:

worldToCamera - 3x4 camera matrix for transforming a 3D point from world to image coordinates.

X - 3D Point in world reference frame..

Returns:

2D Render point on image plane.

splitAssociated

public static void splitAssociated(List<AssociatedPair> pairs,
                   List<georegression.struct.point.Point2D_F64> view1,
                   List<georegression.struct.point.Point2D_F64> view2)

Takes a list of AssociatedPair as input and breaks it up into two lists for each view.

Parameters:

pairs - Input: List of associated pairs.

view1 - Output: List of observations from view 1

view2 - Output: List of observations from view 2

splitAssociated

public static void splitAssociated(List<AssociatedTriple> pairs,
                   List<georegression.struct.point.Point2D_F64> view1,
                   List<georegression.struct.point.Point2D_F64> view2,
                   List<georegression.struct.point.Point2D_F64> view3)

Takes a list of AssociatedTriple as input and breaks it up into three lists for each view.

Parameters:

pairs - Input: List of associated triples.

view1 - Output: List of observations from view 1

view2 - Output: List of observations from view 2

view3 - Output: List of observations from view 3

createCameraMatrix

public static org.ejml.data.DenseMatrix64F createCameraMatrix(org.ejml.data.DenseMatrix64F R,
                                              georegression.struct.point.Vector3D_F64 T,
                                              org.ejml.data.DenseMatrix64F K,
                                              org.ejml.data.DenseMatrix64F ret)

Create a 3x4 camera matrix. For calibrated camera P = [R|T]. For uncalibrated camera it is P = K*[R|T].

Parameters:

R - Rotation matrix. 3x3

T - Translation vector.

K - Optional camera calibration matrix 3x3.

ret - Storage for camera calibration matrix. If null a new instance will be created.

Returns:

Camera calibration matrix.