A different decay behavior is detected in luminescent inorganic materials typically present on cultural heritage objects. For example, semiconductor pigments, such as cadmium- and zinc-based pigments, are typically characterized by a fast picosecond band gap emission, due to the recombination of an ele
Biometrics has recently received significant attention as an alternative to personal authentication methods such as keys, IDs, and passwords [1]. Among various forms of biometrics, face recognition has advantages, such as user acceptance and inexpensive optical sensors. Two-dimensional (2D) face recognition, which uses 2D face images, has dramatically grown in recent decades due to the advancement of computer vision and pattern recognition technologies [2�C4].

Although recent 2D face recognition systems have reached a certain level of maturity under certain conditions, external and internal variations, such as pose, illumination, and expression, continue to affect its overall performance. To alleviate these variations, three-dimensional (3D) face recognition has recently received considerable attention [5�C8].3D face recognition uses 3D face data, which has depth (z) information in addition to the pixel information on (x, y) coordinates of 2D face data. Because 3D face recognition exploits 3D face shape from depth information, which is invariant to external changes, it guarantees better performance than 2D face recognition regardless of external conditions [5�C9]. However, 3D face recognition requires a pose normalization step to make two sets of 3D face data into the same pose, such as a frontal pose.

Moreover, because 3D face data contains additional depth information, 3D face recognition has limitations in terms of memory efficiency and Entinostat computational cost compared to 2D face recognition.The performance of 3D face recognition depends on how precise, noiseless 3D face data is acquired. To acquire accurate 3D face data, a variety of 3D face acquisition systems have been developed [10�C14]. These systems can basically be divided into two categories-active sensing or passive sensing-based on whether or not there are emitting sources. Most existing passive sensing techniques are based on a stereo vision system, which uses multiple images taken by two cameras [10]. Although stereo vision system only requires multiple cameras to reconstruct 3D face data, it must find a set of accurate corresponding points in one image.

These points can be identified as the same points in another image for 3D reconstruction. However, because a face image does not have distinct features except for eyes, nose, and lip regions, it is almost impossible to find precise correspondences according to full-range face images. This is referred to as the correspondence problem of the stereo vision system [15].