LAUSR

The face image analysis field is a well-established research area in computer vision and image processing. An important requirement for accurate face image analysis is a high-quality input face image. In different real-life scenarios, however, the face is often not properly illuminated, which makes the face analysis very difficult or impossible to accomplish. Although a better performance is obtained by changing the spectrum from visible to near-infrared, it is still not enough for extreme illumination conditions. To obtain a high-quality near-infrared face image, a fast automatic brightness control method using approximate face region detection is proposed, which properly adjusts the brightness of the face part of the image. A novel algorithm for approximate face region detection based on spatio-temporal sampled skin detection is proposed together with the split-range feedback controller and the face absence handle. The proposed method is much faster than state-of-the-art solutions and accurate in approximate face region detection. The complete execution time is lower than 10 milliseconds which makes it suitable for hard real-time embedded system implementation and usage, while the reference brightness value is achieved within 10–15 frames, making it robust to extreme illumination conditions in a scene.