Conference Paper

A Linear Approach to Face Shape and Texture Recovery using a 3D Morphable Model.

DOI: 10.5244/C.24.75 Conference: British Machine Vision Conference, BMVC 2010, Aberystwyth, UK, August 31 - September 3, 2010. Proceedings
Source: DBLP
Download full-text


Available from: William A. P. Smith, Oct 17, 2014
10 Reads
  • Source
    • "Hence, despite minimal difference in performance, we fix this number for further experiments, since shape reconstruction in approximately one second is still orders of magnitude faster than the full 3DMM fitting algorithms [3] [26] [27], which need over one minute to fit an image. We now compare our work against state-of-the-art approaches on BFM, namely the basic 3D shape reconstruction Blanz04 [6], self-occlusion handling with visible contour landmarks Qu14 [22] and linear modeling of generalization error Aldrian10 [1]. Additionally, integration of Qu14 into Aldrian10 makes it possible to build a strong baseline, referred to as Aldrian10+Qu14. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Direct reconstruction of 3D face shape—solely based on a sparse set of 2D feature points localized by a facial landmark detector—offers an automatic, efficient and illumination-invariant alternative to the conventional analysis-by-synthesis 3D Morphable Model (3DMM) fitting. In this paper, we propose a novel algorithm that addresses the inconsistent correspondence of 2D and 3D landmarks at the facial contour due to head pose and localization ambiguity along the edge. To facilitate dynamic correspondence while fitting, a small subset of 3D vertices that serves as the contour candidates is annotated offline. During the fitting process, we employ the Levenberg-Marquardt Iterative Closest Point (LM-ICP) algorithm in combination with Distance Transform (DT) within the constrained domain, which allows for fast convergence and robust estimation of 3D face shape against pose variation. Superior evaluation results reported on ground truth 3D face scans over the state-of-the-art demonstrate the efficacy of the proposed method.
    26th British Machine Vision Conference (BMVC), Swansea, UK; 09/2015
  • Source
    • "The key advantage of this approach is that shape and texture estimation can both be posed as multilinear (and hence convex) fitting problems and solved independently. We begin by estimating 3D shape parameters and the camera matrix by fitting to sparse feature points using the algorithm proposed in [1]. This approach uses an empirical model of the generalisation capability of each feature point. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Well known results in inverse rendering show that recovery of unconstrained illumi-nation, texture and reflectance properties from a single image is ill-posed. On the other hand, in the domain of faces linear statistical models have been shown to efficiently characterise variations in face shape and texture. In this paper we show how the inverse rendering process can be constrained using a morphable model of face shape and tex-ture. Starting with a shape estimate recovered using the statistical shape model, we show that the image formation process leads to a system of equations which is multilinear in the unknowns. We are able to estimate diffuse texture, specular reflectance properties, the illumination environment and camera properties from a single image. Our approach uses relaxed assumptions and offers improved performance in comparison to the current state-of-the-art morphable model fitting algorithms.
  • [Show abstract] [Hide abstract]
    ABSTRACT: 3D face shape provides a pose and illumination invariant description of human faces. In this paper, we propose a novel component based method to recover the full 3D face shape from a set of sparse feature points. We use a local linear fitting (LLF) scheme so that reconstruction of each subregion depends on both its own vertices and adjacent subregions. This method results in a separate set of shape coefficients each emphasizing the quality of one subregion and improves the model expressiveness. Experiments show that the LLF strategy significantly reduces the model residual error, and thus reduces the sparse reconstruction error under pose variations. Moreover, the problem of estimating pose parameters is revisited, and we use a joint optimization method to improve the reconstruction quality under unknown pose. We evaluate the sensitivity of our method to the selection of feature points. Simulation results show that our method is more robust than prevailing methods.
    The Visual Computer 02/2014; 30(2). DOI:10.1007/s00371-013-0795-3 · 0.96 Impact Factor
Show more