Fingertips Detection Algorithm Based on Skin Colour Filtering and Distance Transformation

Multi-Fingertip location algorithm is always difficult and hot in Finger-based human-computer interaction systems. There are two major difficulties in this field: 1) obtaining accurate hand binary image, 2) locating fingertips in hand binary image. This article presented a multi-fingertip real-time track and location algorithm based on skin colour and distance transformation. The algorithm consists of the following steps. First, it use elliptical boundary model to detect skin colour of human hand in YCbCr colour spaces. After that, we use distance transformation to filter finger from hand, leaving palm area only. Meanwhile, it uses zero and first moment to calculate center of gravity of palm. Then the initial position of fingertips and finger-roots can be located by pixels of hand edge to the center of gravity of palm. Last, it accurately locates fingertips according to the position relationship between fingertips and finger-roots. Experimental results show fingertips can be located quickly and accurately by the algorithm, which fully meets the requirements of real-time computing tasks.