Max-min hand cropping method for robust hand region extraction in the image-based hand gesture recognition

Abstract

There have been many developments and applications based on hand posture recognition to make human-computer interaction/interfaces more convenient and effective. And, many of these applications are based on the image-processing technique since it can guarantee more information and more flexibility for processing. To develop a hand posture recognition system, the proper extraction of pure hand region is a very important step since it is much related with the final performance and recognition rate. But, by the noisy data due to the illumination, image resolution, and non-uniform distribution of skin colors which could be easily found in real environments, it is not easy to extract the pure hand region exactly. In this research, a simple and effective algorithm for hand cropping, named as max-min hand cropping, is proposed and compared with some of the previous research. Finally, the effectiveness of the proposed method is verified with 152 different hand images from 8 persons and 19 hand postures.