Research And Implement The Interactive Matting Algorithm

With the overall popularity of digital cameras, computers, tablets, smart phones and other devices, our daily life and work is always associated with a large number of multimedia videos, images and other digital media. Hence, matting processing, as the common image processing method, is demanded simpler and more convenient, and with better results. This paper studies the interactive matting algorithm in depth to improve the interactive mode while ensuring the matting quality, and reduce human-computer interaction in matting process. Lazy Snapping algorithm and Grab Cut algorithm are first studied, then Lazy Snapping based on local optimization and Grab Cut algorithm based on significant nature are proposed to overcome certain shortcomings. Finally, experiment results proved that the improved algorithm simplify the interactive mode and improve the prospect extraction effects compared with the original algorithm. The main contribution and contents of this paper are as follows:1、A Lazy Snapping algorithm based on local optimization is proposed. Two aspects are improved based on the original Lazy Snapping algorithm:(1) the simple linear iterative clustering (SLIC) algorithm is used in image pre-segmentation to solve the over-segmentation problem by watershed pre-segmentation in original algorithm.(2)Local optimization is conducted in border areas of prospect segmentation results, which is obtained by marked the foreground and background, then suspicious areas are detected idiomatically and optimized.2、Grab Cut algorithm based on saliency is proposed. Combined with RC significant algorithm and Grab Cut algorithm, this algorithm can automatically get the rectangle included foreground region. The size and position of the rectangle can be changed in order to obtain satisfactory results, then initialize the parameters of mixed Gaussian model and iterate to get foreground area.3、The interactive matting system is achieved. This system realizes the matting function of original Lazy Snapping algorithm, the original Grab Cut algorithm and two improved algorithms proposed in this paper. Finally, an example is used to introduce the use of this system.