You are here: Start » Function Reference » Image » Image Spatial Transforms Maps » CreatePerspectiveMap_Points

CreatePerspectiveMap_Points

Header:	AVL.h
Namespace:	avl
Module:	Calibration

Creates a perspective transform map from four points denoting a rectangle in the world coordinates.

Applications: Data preprocessing for fast perspective correction. The result is used by RemapImage.

Syntax

void avl::CreatePerspectiveMap_Points
(
	const avl::ImageFormat& inImageFormat,
	atl::Optional<const avl::Region&> inRoi,
	const atl::Array<avl::Point2D>& inImagePoints,
	atl::Optional<const atl::Array<avl::Point2D>&> inTargetPoints,
	atl::Optional<const avl::Size&> inNewSize,
	const avl::InterpolationMethod::Type inInterpolationMethod,
	avl::SpatialMap& outSpatialMap,
	avl::Matrix& outTransformMatrix
)

Parameters

Name	Type	Default	Description
inImageFormat	const ImageFormat&
inRoi	Optional<const Region&>	NIL	Range of pixels to be processed
inImagePoints	const Array<Point2D>&		Points define real object plane corners.
inTargetPoints	Optional<const Array<Point2D>&>	NIL	Points define target plane corners. If NIL then image corners are used starting from (0,0) (Width, 0) (Width, Height), (0, Height)
inNewSize	Optional<const Size&>	NIL	New image size after remapping
inInterpolationMethod	const InterpolationMethod::Type		Interpolation method used in extraction of image pixel values
outSpatialMap	SpatialMap&		Created SpatialMap with perspective transform
outTransformMatrix	Matrix&		Used transform matrix

Description

This operation computes a SpatialMap which can be later used for removing a perspective distortion from an image.

The operation maps the input image points (inImagePoints) onto locations described by the target points (inTargetPoints). If the inTargetPoints input is set to Auto this array will be made from the corner points of the input image. Both array must contains four points.

The input inNewSize allows rescaling of the output image.

The inImageFormat format is necessary for preparation of a spatial map.

The outTransformMatrix output allows verifying the found transformation.

Remarks

This filter is a good choice for local perspective distortion removal - such as "unwrapping" boxes, as is depicted by the example above. Applications concerned with observing real-world flat surfaces (such as observing conveyor belts) should use methods that are more accurate. Please refer to: Machine Vision Guide - Camera Calibration and World Coordinates

Notice that both inImagePoints array and inTargetPoints must be made of four points. This filter creates point to point transform so changing the points' order in the array may yield an unexpected result.

The best way to understand the relation between inImagePoints and inTargetPoints is to present the first of them on the input image and the second on the background of the remapped image.

The image before remapping on the left (with inImagePoints) and the image after on the right (with inTargetPoints) created by CreatePerspectiveMap_Path.

Errors

List of possible exceptions: