£. Automatic implementation of U-face
Automatic implementation of U-face is related to face
detection. First detected face, and then in the detected face
region to detect the face contour, pupils of the eyes, the last
auto-calibration of U-face, filling the background.
Face detection requirements a relatively high resolution
face images, which can improve the detection accuracy. Face
region detection can use OpenCV (Open Source Computer
Vision Library); OpenCV is the Intel Corporation to support
the open source computer vision library. It is lightweight and
efficient, able to realize image processing and computer
vision aspects of many common algorithms.
By Visual C++ 6.0 using the OpenCV tested on Figure
2, get the figure 4. We modifY parameters in the program and
the re-detection face Figure 2 get Figure 5, Figure 6 is the
focus area of Figure 5.
We call Figure 6 for the U-2 face model; we treated
Figure 5 and obtained Figure 6 through the face detection
function cvHaarDetectObjects of OpenCV.
If the face image has a certain deflection, we will rotate
the image. If we can detect human face and deflection, and
then rotate the image; if no face is detected, rotate the image
a certain angle and then detected, If still no face is detected,
further rotation angle ... ; until the rotation angle to 360
degrees, we have not found that face, it shows that the image
is not face image.
The histogram normalization is necessary before face
detection.