Citation Context

...yes and the nose. In the 3D face recognition approach used by Faltemier et al. [4], the nose tip and 28 small regions were selected automatically for improving recognition. More recently, Wang et al. =-=[5]-=- computed a signed shape difference map (SSDM) between two aligned 3D faces as a intermediate representation for the shape comparison. Based on the SSDMs, Haar-like, Gabor, and Local Binary Pattern (L...

Citation Context

...only preprocessing applied was a simple median filter on the depth data. In both types we report the VR at FAR=0.1% for the ROC III (the interested reader for more details on the ROC III may refer in =-=[12, 20]-=-). The best results achieved in the ROC III for all the tested methods and the stateof-the-art in 3D face recognition literature are summarized in Table 1 (the results of the other methods are directl...

Citation Context

...rent emphases. Instead of generic 3D objects, McGill benchmark [13] focuses on objects with articulating parts. The partial retrieval problem is raised in [15]. 3D face recognition and identification =-=[16]-=- can been seen as a subproblem of the broad 3DOR. It is also worth noting that the SHREC contest makes a great effort to refine the problem into some sub-problems, such as “mesh soup” vs. watertight m...

Citation Context

...or an improved matching result [5] [6]. Meanwhile, automatic segmentation of facial surface into rigid and mimic regions is still problematic [2]. The second stream builds a generic face model and for a given non-neutral face, a virtual face is generated by this model with reduced facial expression effects. Recognition is then performed on virtual 3D faces [7] [8] [9]. These techniques improve the performance to some extent but they are rather computationally expensive. The last stream proposes to learn the distributions of intra-class and inter-class variations across expression changes [10] [11], and generally leads to better performance. The downside is that it requires a large training dataset with at least several expressive face models for each subject. In this paper, we make the assumption that, when a facial expression occurs, there are always some small local areas that vary slightly or keep invariant as compared to the neutral expression, and these local areas can be found in both relative rigid and elastic facial regions. Once located and characterized, these local regions can be used to achieve 3D face recognition, which is robust to facial expressions and partial occlusion...

Citation Context

... Tan et al. [12] investigated the Lambertian reflectance information, in order to differentiate 3D faces from 2D images. Bao et al. [13] compared the difference of the optical flow fields under various conditions for the same purpose. These methods require high-quality inputs and the computational complexity is very high. Optical flow computations, for example, are 10000 times slower than regular pixel level operations on a video frame [24]. Besides, their performance largely depends on the complexity of the ambient illumination. 3D face recognition has been widely studied in the recent years [14] [15]. It compares the input to a pre-built 3D head template instead of 2D templates. If the face capturing in each authentication trial is also 3D, this approach is robust to 2D media attacks. However, 3D face recognition is still in its infancy. Sandbach et al. [21] presented a survey on 3D face expression recognition and concluded that the high frame rate requirements and frame resolution limit its practical usage. More importantly, the 3D capturing process is much more time consuming than 2D methods or entering a password, therefore it is not suitable for device unlock or app login. We wil...

Citation Context

...covered pattern of asymmetry is itself symmetric. I. INTRODUCTION Facial symmetry is a recurrent topic in computer vision, with several and diverse applications including face recognition [10], [19], =-=[29]-=-, gender classification [30], landmark detection [23], [26], assessment of orthodontic surgery [1], [5] or facial attractiveness [25], to name a few. A particularly interesting application arises from...