Research Lab

 

Face Quality Assessment

Introduction firefox It is often useful to summarize surveillance video by collecting the images of faces that are visible in the original video sequence. These collections, which are called face image logs, allow investigators to determine who was in the vicinity of the surveillance camera at any particular moment in time. The solution resulted from this research strives to construct face image logs that are complete and concise in the sense that the logs contain only the best images available for each individual observed. The method begins by describing how to assess and compare the quality of face images. Then a robust method for selecting high quality images is used, which takes into consideration the limitations inherent in existing face detection and person tracking techniques. Experimental results demonstrate that face logs constructed in this manner generally contain fewer than 5% of all detected faces, yet these faces are of high quality, and they represent all individuals detected in the video sequence. The following figure shows the criteria used to assess the quality of the captured faces.

Pose
Illumination
Sharpness
Skin content
Resolution

Sample frames from the representative test video sequence.
The contents of the face image log after the processing of the representative video.