TIEVisionFaceDetector.alignFace

Declaration

function alignFace(image: TIEVisionImage; face: TIEVisionDetectedFace; size: int32_t = 112): TIEVisionImage; safecall;

Description

Returns an upright, square crop of a face. The image is rotated, scaled and cropped so the eyes, nose and mouth are always at the same positions.
Aligned faces make face recognition (TIEVisionFaceRecognizer) much more reliable, because every training and query image is framed the same way.

Parameter	Description
image	The image that was passed to detect
face	A face returned by detect
size	Width and height of the result, from 16 to 1024. Default: 112

The result has the same pixel format as the source, except that 1 bit images give an 8 bit gray result.
The crop is tight around the face (forehead to chin), the standard framing for face recognition.

Demos

	Demos\IEVision\GetFaces\GetFaces.dpr
	Demos\ImageEditing\EveryMethod\EveryMethod.dpr

Examples

Also see: Automated Samples

// Add every face in the image to a face recognizer, labelled "Natalie"
img := ImageEnView1.IEBitmap.GetIEVisionImage();
faces := detector.detect( img, IEVisionRect( 0, 0, 0, 0 ));
for i := 0 to faces.size() - 1 do
m_faceRecognizer.addTrainImage( detector.alignFace( img, TIEVisionDetectedFace( faces.getObj( i )), 112 ), 'Natalie' );
m_faceRecognizer.train();

// Output the first face found, aligned upright at 112x112
detector := IEVisionLib.createFaceDetector();
img := ImageEnView1.IEBitmap.GetIEVisionImage();
faces := detector.detect( img, IEVisionRect( 0, 0, 0, 0 ));
if faces.size() > 0 then
ImageEnView1.IEBitmap.AssignIEVisionImage( detector.alignFace( img, TIEVisionDetectedFace( faces.getObj( 0 )), 112 ));
ImageEnView1.Update();

// Shows all faces found, side by side

TIEVisionFaceDetector.alignFace