Skip to content

Latest commit

 

History

History
39 lines (25 loc) · 1.19 KB

README.md

File metadata and controls

39 lines (25 loc) · 1.19 KB

Goal

Show virtual objects (hats and glasses) on a face using OpenCV.

Hats and glasses processing

First, we detect face and scale width and height of hat source accordingly.

Then, we resize hat source and hat mask (mask is for transparent background):

cv.resize(hatSrc, hatDst, size);
cv.resize(hatMask, hatMaskDst, size);

After that, we calculate absolute coordinates for hat.

Using these coordinates we copy result image to the main video source src and apply transparent mask:

hatFrames[i].src.copyTo(src.rowRange(hatFrames[i].y1, hatFrames[i].y2)
                           .colRange(hatFrames[i].x1, hatFrames[i].x2),
                        hatFrames[i].mask);

Similar algorithm is used to resize and draw glasses.

Additionaly, we rotate glasses using warpAffine(...) function and rotation matrix:

cv.warpAffine(glassesDst, glassesDst, glasses.rotMat, rSize);
cv.warpAffine(glassesMaskDst, glassesMaskDst, glasses.rotMat, rSize);

References

  1. Face and eyes detection (OpenCV)
  2. Geometric Transformations of Images (OpenCV)