An Improved Watershed Transform Algorithm For Two-Hand Tracking Under Partial Occlusion
Loading...
Date
2016-05
Authors
Leim, Peng Peng
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
To achieve a natural interaction in augmented reality environment, two-handed gesture interactions are highly preferred. However, two-handed interactions always result in mutual occlusions which interfere with the hand gesture recognition. In this research, a solution for this problem is presented by improving the watershed transform algorithm and developing an advanced two-hand tracking system based on vision-based recognition system. The procedure of solving two overlapping hands starts with skin colour detection to acquire the regions of interest which are then assumed to be the candidates of the hands’ regions, followed by the computation of image gradients from a grey colour image to obtain the boundaries of two hands. Next, a seed point is extracted from input image for flooding process. Flooding process is started from the seed point and it ends when it reaches the boundary. The pixels filled in the flooding process are considered as the region of one of the hands. Subtraction method is then applied to extract the second hand region. For hand tracking system development, four possible scenarios of hand tracking are considered: i) no hand, ii) one hand, iii) two separate hands, and iv) two overlapping hands. In the tracking process, we also implemented hand feature extraction to ensure that both output regions are hand region.
Description
Keywords
Improving the watershed transform algorithm and developing an advanced , two-hand tracking system based on vision-based recognition system.