An Integrated Approach: Combining GrabCut and Contour-Matching for Hand Gesture Segmentation in Indian Sign Language

Hand Segmentation plays a vital role in human-computer interaction using computer vision. It serves as an initial step in hand gesture recognition systems or to recognize various sign languages and is considered essential pre-processing. The region of interest (ROI) in an image requires filtering or modification, which can be represented as a bounding box or a binary mask picture. Accurate hand segmentation is the critical first phase in sign language recognition (SLR) systems. Segmenting the ROI reduces processing time and enhances the precision of sign recognition. This study proposes two methods: a contour-matching technique applied to a simple black background dataset, and a novel hybrid algorithm that combines the GrabCut method with the contour-matching method to extract the ROI from images with complex backgrounds. This work’s purpose is to use the segmented images for the Indian sign language (ISL) recognition system. The results show that the contour-based segmentation technique achieves excellent results for the dataset with a black background. The number of iterations the GrabCut algorithm needs to separate the foreground varies depending on the complexity of the background. The algorithms are evaluated on datasets of ISL consisting of images with a black background and images with a complex background.