A Cluster Sampling Method for Image Matting via Sparse Coding
In this paper, we present a new image matting algorithm which solves two major problems encountered by previous sampling-based algorithms. The first is that existing sampling-based approaches typically rely on certain spatial assumptions in collecting sam
- PDF / 6,556,588 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 86 Downloads / 292 Views
Abstract. In this paper, we present a new image matting algorithm which solves two major problems encountered by previous samplingbased algorithms. The first is that existing sampling-based approaches typically rely on certain spatial assumptions in collecting samples from known regions, and thus their performance deteriorates if the underlying assumptions are not satisfied. Here, we propose a method that a more representative set of samples is collected so as not to miss out true samples. This is accomplished by clustering the foreground and background pixels and collecting samples from each of the clusters. The second problem is that the quality of matting result is determined by the goodness of a single sample pair which causes errors when sampling-based methods fail to select the best pairs. In this paper, we derive a new objective function for directly obtaining the estimation of the alpha matte from a bunch of samples. Comparison on a standard benchmark dataset demonstrates that the proposed approach generates more robust and accurate alpha matte than state-of-the-art methods. Keywords: Image matting Foreground extraction
1
·
Sampling
·
Clustering
·
Sparse coding
·
Introduction
Estimation of foreground and background layers of an image is fundamental in image and video editing. In the computer vision literatures, this problem is known as image matting or alpha matting. Mathematically, the process is modeled in [1] by considering the observed color of a pixel as a combination of foreground color and background color: Iz = αz Fz + (1 − αz Bz )
(1)
where Fz and Bz are the foreground and background colors of pixel z, αz represents the opacity of a pixel and takes values in the range [0,1] with αz = 1 for foreground pixels and αz = 0 for background pixels. This is a highly ill-posed problem since we have to estimate seven unknowns from three composition equations for each pixel - one for each color channel. Typically, matting approaches c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part II, LNCS 9906, pp. 204–219, 2016. DOI: 10.1007/978-3-319-46475-6 13
A Cluster Sampling Method for Image Matting via Sparse Coding
205
rely on constraints such as assumption on image statistics [2,3] or user interactions such as a trimap to reduce the solution space. A trimap [4] partitions an image into three regions - known foreground, known background and unknown regions that consist of a mixture of foreground and background colors. From the aspect of assumptions on image statistics, existing natural image matting methods fall into three categories: (1) propagation-based [2,5–10]; (2) color sampling-based [11–18]; (3) combination of sampling-based and propagation-based [19–22] methods. Propagation-based methods assume that neighboring pixels are correlated under some image statistics and use their affinities to propagate alpha values of known regions toward unknown ones. Samplingbased methods assume that the foreground and background colors of an unknown pixel can be explicitly estim
Data Loading...
