An improved watershed transform algorithm for two-hand tracking under partial occlusion
Leim Peng Peng, Tan Guat Yew · AIP conference proceedings · 2017
This article presents an improved watershed transform algorithm to extract two hand region from an image. The original algorithm of watershed transform is vulnerable to noise, which may yield miscellaneous over-segmentation where a large number of segmented regions are formed. To segment two hand region from an image, we suggest to use image gradient computational method from a grey color image to obtain the boundaries of two hands. Next, seed-fill algorithm is used to detect one of the hands in the image. In this algorithm, a seed point is extracted from input image for flooding process, which starts from the seed point and ends when reaching the image boundary. Subtraction method is then applied to extract the region of the other hand. In the tracking process, we also implemented hand feature extraction to ensure that both output regions are hand region. By using this improved watershed algorithm, two hand region can be extracted directly from region of interest of an image. In contrast, the result obta...