Image matting trimap optimization by ant colony algorithm
- PDF / 5,426,789 Bytes
- 27 Pages / 439.642 x 666.49 pts Page_size
- 47 Downloads / 233 Views
Image matting trimap optimization by ant colony algorithm Genji Yuan1,2 · Jinjiang Li1,2
· Zhen Hua3
Received: 19 March 2020 / Revised: 21 August 2020 / Accepted: 16 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract In this paper, we present a way to create an accurate trimap. Since the matting problem is a serious under-constrained problem, the user is required to provide additional constraint information to estimate the alpha value of the mixed pixels. The smaller the area of the unknown region of the user-provided trimap, the more accurate the alpha value of the estimated mixed pixels. But manually creating the trimap is a complicated and time-consuming task, and it is even impractical to manually create trimaps for some tasks. We use the ant colony algorithm to determine the boundary information of the foreground objects, and fuse different pheromone images at the superpixel level to create an accurate trimap. The trimap generated by our method also has higher precision when the edge of the foreground object has a lot of fine hair. Experiments show that the high-quality trimap can be generated by the method of this paper, which can effectively improve the performance of the matting algorithm and achieve accurate alpha mask estimation. Keywords Matting · Trimap · Ant colony · Co-fusion
1 Introduction The process of alpha matting is to separate the foreground information and background information in an image, which is an important issue in the field of image processing and Jinjiang Li
[email protected] Genji Yuan [email protected] Zhen Hua [email protected] 1
School of Computer Science and Technology, Shandong Technology and Business University, Yantai 264005 China
2
Co-innovation Center of Shandong Colleges and Universities: Future Intelligent Computing, Yantai 264005 China
3
School of Information and Electronic Engineering, Shandong Technology and Business University, Yantai 264005 China
Multimedia Tools and Applications
video editing and widely used in the field of image and video editing. The mathematical model of alpha matting is as follows: I = αF + (1 − α) B,
(1)
where α is opacity and represents the linear mixture of foreground color and background color. F, B represent foreground pixel and background pixel respectively. An image can be viewed as a combination of foreground objects and background information. α can take a value in [0, 1], indicating a linear combination of foreground and background. If α = 1, the pixel belongs to the foreground completely; if α = 0, the pixel belongs to the background completely. In most natural images, most pixels belong to absolute foreground or absolute background, and accurate estimation of the alpha value of mixed pixels is the key to solving matting problem. Given an input image, for all pixels, its (F, B, α) is unknown and needs to be estimated. Considering that the natural images we are dealing with are mostly RGB images, we can obtain three equations (one for each channel) and seven unknowns
Data Loading...
