Exemplar-Based Graph Matching for Robust Facial Landmark Localization

摘要: Localizing facial landmarks is a fundamental step in image analysis. However, the problem still challenging due to large variability pose and appearance, existence of occlusions real-world face images. In this paper, we present exemplar-based graph matching (EGM), robust framework for landmark localization. Compared conventional algorithms, EGM has three advantages: (1) an affine-invariant shape constraint learned online from similar exemplars better adapt test face, (2) optimal configuration can be directly obtained by solving with constraint, (3) optimized efficiently linear programming. To our best knowledge, first attempt apply technique Experiments on several datasets demonstrate advantages over state-of-the-art methods.