The Research On Hand Gestures Segmentation Based On Difference Background Image

Compared with other human-computer interactive modes, hand gesture possesses the characteristics of naturalness, simplicity, directness and richness. During the process of hand gesture recognition, the quality of gesture segmentation directly influences the results of hand tracking, gesture feature extraction and gesture recognition in follow-up works.In the research of hand gesture segmentation based on visual, in order to capture and control the source of hand gesture video, the visual studio 2003 and the DirectShow are used as the development tools after studying and analyzing the basic theories and methods of video capture. Experimental software based on the combination of WDM and DirectShow is designed and implemented. The software realized the following functions: controlling and setting the properties of video source, capturing the video source, saving the video sequences, outputting the single hand image or a series of hand images and video files, generating key frames for hand gesture analyzing in real time.After studying methods and basic theories of hand gesture segmentation and recognition thoroughly, the current research status is summarized and various methods of segmentation are analyzed. A method of hand gesture segmentation, which is based on the background difference, is studied and implemented in experimentation. Based on the experimentation, the features of the background noise and the gesture shadows is found out and a new method of updating the background dynamically is proposed to eliminate the effects of background noise and the gesture shadows. After many experiments, experimental results show that: the new method of updating background can be effective to reduce factors of background noise and gesture shadows.Finally, an adaptive-threshold segmentation method based on the block is proposed to segment the hand gesture images and the image morphological method is used to process the hand image and to output the binary images of hand gesture. In order to verify the new method proposed, the segmentation effects using the Otsu method with static background and updating background are compared with the new method respectively. The experimental results show that: the new method proposed can generate the ideal hand gesture images for following works.