Published in the proceedings of “35 th Annual International Conference of the IEEE Engineering in Medicine and Biology Society”, Osaka, Japan, July 3-7, 2013 Abstract— Accurate vascular pattern localization has many applications in the diverse scientific and application domains. For example, vascular patterns not only have been widely used as a biometric-based human identification method that is inexpensive, secure and easy to use, but also have produced more accurate heart-rate estimation using conventional RGB camera by defining regions of interest along the vascular patterns instead of the entire exposed skin area. In addition, extracting temporal activity along vascular patterns can further enable targeted monitoring of other physiological parameters, such as blood flow and blood pulse transition time. This paper presents a method for robust and accurate vascular pattern localization using conventional RGB imaging systems. Our approach overcomes current limitations of systems that use still RGB images for vascular pathway localization - which produce low contrast between areas of vascular patterns and skin tissues and are sensitive to skin color variations - by capturing the temporal differences between these two areas from RGB videos. I. BACKGROUND AND INTRODUCTION Vascular patterns are the subcutaneous patterns of blood vessels beneath the skin. Vascular patterns are unique to an individual and are very stable over a long period of time [1]. Recently, vascular pattern biometrics has attracted increasing interest from both research communities and industries due to its ease of use, low cost, accuracy and security [2]. Another area where vascular pattern localization is bringing value is in providing more accurate heart-rate (HR) estimation using conventional RGB camera by defining regions of interest for sensing along vascular patterns instead of sensing across a broad area of exposed skin, especially in the back of the hand [3]. Previous work has shown that infrared imaging provides a noncontact and noninvasive data acquisition method for capturing superficial vascular patterns and does not require injection of any agents into the blood vessels [1, 4-12]. Therefore, it is by far the best known noninvasive technology to acquire vascular pattern images. L. Wang and L. Graham [8] compared near- and far-infrared imaging for capturing vascular patterns and concluded that the near-infrared (NIR) imaging produces good quality images when capturing vein patterns in the back of the hand, palm, and wrist and it is more tolerant to changes in environmental and body condition. NIR imaging for vascular pattern localization relies on two properties of the infrared radiation: 1) light in 700 to B. Xu, L. K. Mestha, and Robert. P. Loce are with Xerox Corporation, Webster, NY 14580, USA (corresponding author phone: 585-422-5768; fax: 585-422-6117; e-mail: Beilei.Xu@xerox.com). Y. Liu is with Massachusetts Institute of Technology, Cambridge, MA 77005 USA (e-mail: yiliudd@mit.edu). 900 nm spectral window penetrates sufficiently deep into tissues, thus allowing for noninvasive imaging [13]; 2) absorption differences between the reduced hemoglobin and other skin tissues in the NIR spectral band [1]. The absorption difference produces images with high contrast between vascular patterns and skin tissues as shown in Fig. 1(a) and is less sensitive to skin color variations. From Figure 1(a), the vascular pattern can be easily extracted using this high image contrast (e.g., apply a lightness threshold based on image contrast) as shown in Fig. 1(b). (a) (b) Figure 1: (a) An NIR image of the hand; (b) Extracted vascular pattern based on the NIR image. However, NIR-based imaging systems often require special illumination sources and / or specialized NIR cameras in the wavelength range that is suitable for vein pattern localization. These requirements can greatly increase the cost of the imaging system and add implementation challenges. Conventional RGB imaging applied to vascular pattern localization could overcome these costs and convenience limitations. However, in the visible spectral band, as can be seen in Fig. 1(c), the contrast between vascular patterns and skin tissues is much lower than what NIR imaging systems provide (e.g., compare to the hand image in Fig. 1(a)). Because of the low image contrast, the extracted vascular patterns from the analysis of the still image in Fig. 1(c) can only reveal portions of the main vascular patterns as shown in Fig. 1(d). Another factor that makes it difficult to use still RGB images alone is its sensitivity to skin color, where the contrast between vascular patterns and skin tissues can vary greatly depending on the skin color. (c) (d) Figure 1: (c) RGB image of the hand; (d) Extracted vascular pattern based on a single RGB image (threshold based on color). On the other hand, RGB-video imaging enables extraction of the temporal signals which are not available from still images. For example, Wu et. al. was able to reveal the subtle motion of blood vessels arising from blood flow in Vascular Pattern Localization via Temporal Signature Beilei Xu Member, IEEE, Yi Liu, Lalit K. Mestha, Fellow, IEEE, and Robert. P. Loce, Senior Member, IEEE