Image Matting Research Based On Energy And Sampling

Image matting is a key technology in film and media production. It plays an important role in image editing and computer vision. Image matting has a wide application in special effects, virtual reality, augmented reality, home entertainment, and photo manipulation.Image matting is to solve the problem of extracting the foreground object from an image. According to the difference of the background, it divided image matting into Blue Screen Matting and Natural Image Matting. Blue Screen Matting has a close relationship with the film and television production, and it requires the background’s color is blue or green. Blue Screen Matting is relatively simple, and its typical algorithms are color-difference technique and MiShima method. Natural Image Matting has lower requirements of the background image, so it allows any natural image to be the background. Natural Image Matting is the process of extracting a foreground object from an image by estimating a color and opacity for the unknown pixels. Its typical algorithms are Knockout method, Ruzon-Tomasi method, Hillman method, Bayesian Matting, Poisson Matting, and Closed-Form Solution.We have realized a global sampling method for alpha matting. Firstly, users have to provide the trimap. Then we extract all known foreground and background pixels on the unknown region’s boundary; use these samples to generate the FB search space. Finally, it obtains the matte by random searching the best sample pair which minimizes the cost function. The global sampling method can get good results fast. But the results rely on the given trimap, so we present an easy interactive matting method based on the sampling search matting. Users only require drawing some strokes or points on the foreground and background to extract the foreground object. This method can easily collect the samples available to compute the alpha value for the whole image pixels.