Quantitative evaluation of Segmentation algorithms based on level set method for ISL datasets

摘要

The enormous potential research efforts have been taken for sophisticated and natural human computer interaction using gestures. This work has got motivated from long ago as 1980?s since sign language is the only communicate mode for deaf community people. In signing, the face and a hand of a signer often overlap, thus the system needs to segment these for the purpose of feature extraction. This research work concentrates with the separation of the face and hand by detecting contour of the static object using reference labels and different snake algorithms. Indian sign language dataset is used to evaluate a few level set computer vision algorithms. Specifically different feature sets, segmentation algorithms and color constancy algorithms are evaluated quantitatively. In future, it is possible to evaluate the accuracy of sign on a large scale due to the availability of large annotated databases.