**Summary:** The paper introduces GrabCut, an interactive foreground extraction method using iterative graph cuts. It improves upon previous segmentation techniques by combining texture and edge information, reducing user interaction, and introducing border matting for accurate alpha matte estimation. The method uses an iterative optimization approach to refine segmentation and allows incomplete user labeling, making it more efficient. It also incorporates a robust border matting algorithm to estimate alpha values around object boundaries, handling blur and mixed pixels effectively. The core of GrabCut is based on graph-cut optimization, which is enhanced with iterative estimation and incomplete labeling. This allows users to simply drag a rectangle around the object, significantly reducing the need for manual pixel editing. The system first performs hard segmentation using graph cuts, then applies border matting to refine the alpha matte. The border matting process uses a soft step function to estimate alpha values along the boundary, ensuring smooth transitions and avoiding color bleeding from the background. The method is tested against other segmentation tools like Magic Wand, Intelligent Scissors, Bayes Matting, and Knockout 2. It outperforms these in terms of user interaction and result quality, especially in complex scenarios involving camouflage and low-contrast regions. The algorithm is efficient, with acceptable runtime for interactive use, and has been shown to handle moderately difficult alpha mattes effectively. Key contributions include the use of Gaussian Mixture Models (GMMs) for color data modeling, iterative energy minimization for improved segmentation, and border matting for accurate alpha matte estimation. The system is robust, user-friendly, and effective in a wide range of image editing tasks.**Summary:** The paper introduces GrabCut, an interactive foreground extraction method using iterative graph cuts. It improves upon previous segmentation techniques by combining texture and edge information, reducing user interaction, and introducing border matting for accurate alpha matte estimation. The method uses an iterative optimization approach to refine segmentation and allows incomplete user labeling, making it more efficient. It also incorporates a robust border matting algorithm to estimate alpha values around object boundaries, handling blur and mixed pixels effectively. The core of GrabCut is based on graph-cut optimization, which is enhanced with iterative estimation and incomplete labeling. This allows users to simply drag a rectangle around the object, significantly reducing the need for manual pixel editing. The system first performs hard segmentation using graph cuts, then applies border matting to refine the alpha matte. The border matting process uses a soft step function to estimate alpha values along the boundary, ensuring smooth transitions and avoiding color bleeding from the background. The method is tested against other segmentation tools like Magic Wand, Intelligent Scissors, Bayes Matting, and Knockout 2. It outperforms these in terms of user interaction and result quality, especially in complex scenarios involving camouflage and low-contrast regions. The algorithm is efficient, with acceptable runtime for interactive use, and has been shown to handle moderately difficult alpha mattes effectively. Key contributions include the use of Gaussian Mixture Models (GMMs) for color data modeling, iterative energy minimization for improved segmentation, and border matting for accurate alpha matte estimation. The system is robust, user-friendly, and effective in a wide range of image editing tasks.