Why Is Facial Occlusion a Challenging Problem?

Ekenel, Hazım Kemal; Stiefelhagen, Rainer

doi:10.1007/978-3-642-01793-3_31

Cited by 66 publications

(42 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…During each session 13 pictures per person were taken under the same conditions. Similar to the work of [10], only a subset of 110 individuals for which all variability is available is used in our experiments. CMU-PIE.…”

Section: Resultsmentioning

confidence: 99%

“…Several methods for face detection and cropping [33], eye detection and subsequent rectification have been proposed. Using these methods, basic global transformations can be coped with, but it was shown that in addition strong features must be used in matching algorithms [10]. Matching algorithms try to find a matching local descriptor in the reference image for each item in a sparse or dense set of local descriptors representing the test image.…”

Section: Problem Statementmentioning

confidence: 99%

“…Since they are all outperformed even by the ZOW despite using zero-order-like matching algorithms themselves, it can be concluded that the SIFT descriptor in combination with occlusion modelling already provides a significant advantage. 10.54 -DCT [10] 3.59 -Partial Dist. [30] * 4.67 -Stringface [4] 13.00 -PWCMr [14] -16.00 LGBPHS [36] 16.00 -SOM [29] 25 - * used only a subset of occlusions Partial Dist.…”

Section: A Ar-face Occlusionsmentioning

confidence: 99%

“…Here, [10] uses DCT features extracted from non-overlapping blocks, LGBPHS [36] use local Gabor binary pattern histograms and [7] evaluates both SURF and SIFT features using a locally restricted matching scheme. Another two non-learning approaches, namely Stringface [4] and Partial Distance [30], employ matching procedure for recognition and hence are similar to our method: the former one is inspired by a string-based matching after representing a face as an attribute string, while the latter one uses nonmetric partial similarity measure.…”

Section: A Ar-face Occlusionsmentioning

confidence: 99%

“…Since this is not always possible, we propose to use two-dimensional warping (2DW) algorithms in order to obtain a deformation-invariant dissimilarity measure which can be used for nearest-neighbour classification. In contrast to related to feature matching approaches [10,27,5,35,13,7], our approach uses dense information and incorporates structural dependencies. This leads to smooth deformations, which can handle crucial appearance changes resulting from different poses and expressions (c.f.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Warp that smile on your face: Optimal and smooth deformations for face recognition

Gass

Pishchulin

Dreuw

et al. 2011

Face and Gesture 2011

View full text Add to dashboard Cite

Abstract-In this work, we present novel warping algorithms for full 2D pixel-grid deformations for face recognition.Due to high variation in face appearance, face recognition is considered a very difficult task, especially if only a single reference image, for example a mug-shot, per face is available.Usually model-based approaches with additional training data are used to cope with several types of variation occurring in facial imaging. Image warping contrarily yields a distance measure which is invariant with regard to several types of variation. This allows for precise recognition even using only very few reference observations. Due to the computationally complex problem of optimal 2D warping, pseudo-2D warpingbased approaches in the past represented strong approximations of the original problem, and were mainly successful on data with low variability or rectified images.We propose a novel 2D warping method which is globally optimal and makes no prior assumtions on the data variability besides two-dimensional smootheness constraints which both avoid local mirroring and gaps and significantly speed up the optimization. Furthermore, we show that occlusion handling is imperative to obtain smooth warpings in a variety of domains.We evaluate our novel algorithm on various well known databases, such as the AR-Face and CMU-PIE database, and provide a detailed comparison to existing warping approaches. We show that by using simple relative 2D constraints, strong local features and a kernel, which is robust w.r.t. occlusions, our computationally complex approaches outperform state-ofthe-art results for recognizing faces under varying expressions, occlusions and poses. Most interestingly, we achieve higher accuracy using fewer training instances per class compared to methods learning a model of the 3D shape.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Problem Statementmentioning

confidence: 99%