Image matting aims to extract a foreground object from a single natural image by recovering the partial transparency and corresponding color of the foreground object at each pixel in the image. The resulting transparency map is thereby denoted as alpha matte. The matting problem is severely ill-posed, and in this thesis we focus on matting approaches that utilize user interaction to make the problem tractable. There are three fundamental challenges in interactive image matting research that are addressed in this thesis: (i) Providing a fast and intuitive user interface; (ii) finding a good cost function for matting; and (iii) providing a benchmark that allows a quantitative comparison of matting results. In most previous approaches the user interacts with the algorithm by drawing an accurate trimap, which is a partition of the image into foreground, background and unknown regions. An accurate trimap is very tedious to create manually, hence we follow recent work and aim to automatically generate a trimap from very little user input. The novelty of our approach lies in a new cost function that describes the goodness of a trimap solution. Our cost function considers several image cues and incorporates four different types of priors that are used to regularize the result. We show that our method is fast and produces accurate results. Given a trimap, the thesis then addresses the problem of extracting an alpha matte from a single photograph. We improve on previous image matting approaches by assuming that the majority of partial transparencies are induced by the imaging process. Hence we exploit a model where alpha is the convolution of a binary segmentation with the camera´s point spread function. Based on this model, we propose new matting algorithms that generate high-quality results even for images where our assumption is not met completely. Finally, we introduce a new benchmark test for image matting that enables a quantitative comparison of matting results. Our contributions are (i) a challenging, high-quality ground truth test set that builds the basis of our evaluation; (ii) a dynamic online benchmark system that allows other researchers to interactively analyze recent matting work and to complement the evaluation with new results; and (iii) perceptually motivated error metrics for image matting. We use this benchmark to confirm that our proposed matting algorithms outperform the current state-of-the-art.
C. Rhemann: "Interactive Image Matting"; Supervisor, Reviewer: M. Gelautz, C. Rother; 188/2, 2010.
Click into the text area and press Ctrl+A/Ctrl+C or ⌘+A/⌘+C to copy the BibTeX into your clipboard… or download the BibTeX.