Research Of Fast Interactive Image Segmentation Algorithm Based On Grab Cut

It is no doubt that image segmentation is a key technique in computer vision and image processing engineering for the segmenting results will directly feedback to the subsequent image processing steps. There are various segmenting methods, different application possesses its own optimal method, so no general method has been found presently. Grab Cut has got wide attention widely for its convenient interactive operation and can acquire globally optimal solution. Grab Cut algorithm is discussed based on graph theory in this paper. Two methods have been proposed to deal the problem in the traditional Grab Cut that it is hard to consider the real time segmentation and accuracy while dealing with image of large resolution.(1) To solve the difficulty to ensure both accuracy and real-time interaction in Grab Cut resulting from the massive pixels and iterative updated Gaussian mixture model parameters, a fast interactive segmentation method based on super-pixel is proposed in this paper. The method reduces the scale of the problem while keep accurate and real-time segmentation simultaneously. It is achieved through the following four steps. Firstly, it uses Mean Shift algorithm merged with edge confidence to segment the image into some homogeneous regions which maintain the color information, spatial information as well as the feature of boundaries; secondly, it constructs the foreground and background GMM with the mean RBG of the homogeneous regions as the network nodes; then, it combines EM algorithm with Max Flow/Min Cut to estimate the real GMM parameters; finally, it automatically adjusts the weights of the region term in order to improve the segmentation accuracy. The experiment results show that the algorithm presented can response to user interaction timely and it’s useful in dealing with some image with simple foreground and background color distributions, especially make a comprehensive consideration of better identify weak and sharp edge.(2) Three drawbacks are found in the traditional grab cut, which are serious overlaps of foreground and background colors, unsatisfactory result of segmenting single object from multiply objects as well as shrinking bias phenomenon. To solve these drawbacks, Grab Cut algorithm based on polygon user interaction driven is proposed, A fast interactive segmentation framework from global to local based on super pixel is built. First of all, user provides a rough initial contour through a polygon so that more accurate foreground and background seeds through expansion operation can be acquired; Secondly, the Mean Shift algorithm merged with edge confidence is used to pre-segment the image into super-pixels which maintain the color, space and boundary information to provide the possibility for real time interaction. Finally, the energy function is modified through combining the introduced shape information with the interaction term to guide the training of GMM parameters, enrich the characteristics of each super pixel and keep both integrity and smooth edges. The improvements have been proposed aim at solving the defects of traditional Grab Cut as well as improving the time cost, extracting the object of interest quickly and accurately.