Image Matting Based On Deep Learning And Image Dissection
Lei Liu, Yingyun Yang · 2019
Alpha matting is an image processing task designed to separate foreground and background images. Lots of solutions were proposed to attach it. However, there are few matting solutions by deep learning. In this paper, a matting algorithm was proposed, which was based on deep learning and ROF model dissection. We established an end-to-end network to get the alpha matte. In addition, the image was dissected into two parts and put in the network, to get a good result in matting task. The network is based on VGG16 and the images was dissected into cartoons and textures, which brought an excellent result.