Face Algorithms
Face
face detection ⇒ \Rightarrow ⇒ face alignment (marklands) → \rightarrow → face embedding → \rightarrow → face identification
Face Quality Assessment
A binary classification to judge the picture quality is good or not as a face in quality. consider the reduction of computation and increase of reliability
face detection
→
\rightarrow
→ face alignment
→
\rightarrow
→ quality assessment
sub-measurement
- symmetry: to make sure the pose of face satisfied. estimate by histogram intersection distance of left and right of facial local features
- head rotation/pose: measure by center of face and center of picture
- sharpness: to estimate the affect of motion blur. estimate by Laplace operator
- illumination: estimate by the length R of available range of gray intensities excluding 5% of the darkest and brightest pixels
- face size: estimate by size of bounding box
- presence of skin pixels:
- openness of eyes and red eyes:
reference
https://blog.csdn.net/u013082989/article/details/83537370