Acquisition of a Face Database for Video Surveillance Research

In this paper we describe a new database of human faces collected using commercially available video surveillance system, which included 6 surveillance cameras, professional video surveillance recording device, professional photographer´s camera, and DVCAM digital camera. Our new face database consists of the following: 2990 face images, 130 subjects (130 different persons), 23 images per each subject (15 images taken with 5 surveillance cameras at three different distances; 6 images taken with 2 night-vision surveillance cameras at these three distances: 2 mug shot images taken with professional photographer´s camera and night-vision camera). Besides, for each subject we have uncompressed face images at 9 different poses (from -90 degrees to +90 degrees, with the step of 22.5 degrees), and corresponding video sequence with subject´s head moving from left to right and from bottom to top. Our new face database, according to our knowledge, is the first video surveillance face database. This database can be successfully used in different face recognition research areas