Professional Documents
Culture Documents
Abstract—Over-segmentation of an image into superpixels has [6], image matching [7], facial images analysis [8], real-time
become a useful tool for solving various problems in image attention for robotic vision [9], tumor segmentation in medical
processing and computer vision. Reflection symmetry is quite images [10], 3D reconstruction [11], [12], shape manipulation
arXiv:1805.09232v2 [cs.CV] 10 Aug 2018
Fig. 1. (a) The results for the Manifold SLIC [14], (b) results for the proposed algorithm SymmSLIC, and (c) zoomed mirror symmetric windows for both
the methods (top: MSLIC, bottom: ours). We observe that the symmetry at the superpixel level is preserved in ours.
and symmetry axis detection. In Section III-A, we discuss the to get structure sensitive superpixels, where authors perform
proposed method for the detection of mirror symmetry point the SLIC on a 2-dimensional manifold [14]. Liu et al. used the
pairs. In Section III-B, we develop the SymmSLIC algorithm. entropy rate for homogeneous and compact superpixels [28].
In Section IV, we present the results for reflection symmetry Reflection Symmetry. The problem of detecting reflection
aware superpixel segmentation. In Section V-B, we provide symmetry present in images have been thoroughly studied
two applications: unsupervised symmetric object segmentation recently [29], [13]. The existing approaches for symmetry de-
and symmetry axes detection. In Section VI, we conclude the tection in images can be categorized in four categories - direct
paper with discussion on limitations and future directions. approach [30], [31], voting based approaches [32], [33], [34],
[35], [36], basis function based approaches [37], and moment
based approaches [38], [39]. Loy and Eklundh mirrored the
II. R ELATED W ORKS
scale invariant feature transform (SIFT [40]) descriptors in
Superpixel segmentation and symmetry detection problems order to get the reflection invariant SIFT descriptors [40].
have been studied thoroughly and are active research problems In order to match two points they used SIFT descriptor for
in image processing, computer vision, and computer graphics. one point and mirrored-SIFT descriptor for the other point.
To the best of our knowledge, there have not been any previous Then, they detect the symmetry axis using Hough transform
attempt on the problem of symmetry preserving superpixel based line detection algorithm [35]. Kondra et al. proposed
segmentation. We discuss the state-of-the-art methods for a kernel based approach [41]. Patraucean et al. used affine
superpixel segmentation and symmetry detection methods. invariant edge features and a contrarion validation scheme
Superpixels. There are two major categories of algorithms for [42]. Michaelsen et al. used Gestalt algebra [43]. Atadjanov
superpixel segmentation - graph based and clustering based. and Lee detected symmetry axes using appearance of structure
Following are the major graph based approaches. Shi and features [36]. The works [44], [45], [46], [47], [48], [49], [13],
Malik proposed normalized cut algorithm to over-segment [50], [51], [52], [53], [54] present good methods for symmetry
an image [19]. Felzenszwalb proposed a graph-based image detection. There also have been works in detecting symmetry
segmentation approach [20]. Li and Chen used linear spectral in 3D geometric models [55].
clustering approach [21]. [22] and [23] used optimization
techniques for superpixel segmentation. Zhang et al. proposed
a boolean optimization framework for superpixel segmentation III. P ROPOSED A PPROACH
[24]. Moore et al. posed the problem of superpixel detec-
A. Approximate and Partial Reflection Symmetry Detection
tion as lattice detection [25]. Duan and Lafarge used shape
anchoring techniques on the set of detect line segments in Let I : W × H → R3 be a color image with width w and
order to partition the image into convex superpixels [26]. height h, where W = {1, 2, . . . , w} and H = {1, 2, . . . , h}.
The key idea in the clustering based approaches is to first Most of the real images exhibit only the partial reflection
initialize some cluster centers and then refine these cluster symmetry, which means that the mirror symmetric pixel exists
centers using various techniques. Achanta et al. proposed a k- for only a fraction of pixels. This is due to the fact that
means clustering based approach called simple linear iterative the images have square boundaries and the boundaries of
clustering (SLIC) [17]. They initialize cluster centers at the the real object are not necessarily square and furthermore,
centers of equally spaced squares. They perform clustering there could be occlusions and missing parts. Since the real
by assigning each pixel to the nearest center based on the objects are not perfectly mirror symmetric, we attempt to
color and location similarity. Levinshtein et al. proposed a detect the approximate reflection symmetry. Our goal is to
geometric flow based approach [27]. Wang et al. proposed detect the partial and approximate reflection symmetry present
a content sensitive superpixel segmentation approach where in the input image I. We detect the pairs of pixels which are
the distance between the cluster center and a pixel is the mirror reflections of each other. We represent the reflection
geodesic distance [16]. Liu et al. proposed a fast algorithm symmetry present in the image by two subsets, L ⊂ W × H
3
and R ⊂ W × H, satisfying the following property. For each Rjj 0 QR> jj 0 (ci (α) − tjj 0 ) − tjj 0 be the reflection of the curve
pixel xi ∈ L, ∃ xi0 ∈ R such that through the symmetry axis Lj and η i0 j (α) be its normal at α.
If the symmetry axes Li and Lj are similar then η i0 j and η j 0 i
xi0 = Rii0 QR> ii0 (xi − tii0 ) + tii0 , and I(xi ) = I(xi0 ). will also be similar. Therefore, we create an edge between the
R1
vertices vi and vj if 0 (η > >
i0 j (α)η i0 (α) + η j 0 i (α)η j 0 (α))dα >
xi +xi0 1 0
Here, the point tii =0 , the matrix Q = , 2τ .
2
0 −1
cos θii0 − sin θii0 In Figure 4 (a), we show a graphical illustration. We observe
and the matrix Rii0 = . The angle θii0 is that each clique in the graph G corresponds to the set of pairs
sin θii0 cos θii0
the slope of the symmetry axis which is a line perpendicular of mirror symmetric pixels belonging to a same symmetry
to the vector xi − xi0 and passes through the mid-point tii0 . axis [57]. A clique in the graph G is a subset C of the vertex
In order to determine the sets L and R, representing the set V such that every vertex in C is connected by an edge.
reflection symmetry, we use normals of the edges present in Therefore, our goal is to find all the dominant cliques in the
the image. We first extract the edges from the image using [56] graph G. It is a well known result that a clique is equivalent to
and represent them as curves. Let E : W × H → {0, 1} be an independent set in the complement graph. The complement
the image representing the edges present in the image I. Let graph Ḡ = (V̄, Ē) of a graph G is the graph such that V̄ = V
F = {x : E(x) = 1} be the set of pixels lying on the edges. , (u, v) ∈ E ⇒ (u, v) ∈ / Ē, and (u, v) ∈ / E ⇒ (u, v) ∈ Ē. An
Now, for each pixel xi ∈ F, we extract an edge of length p independent set in the graph Ḡ is a subset I of the vertex set
pixels passing through the pixel xi such that the pixel xi lies at V̄ such that no two vertices in I are adjacent. Furthermore, the
equal distance from the end points of the edge, and represent independent set is complement of the vertex cover. A vertex
it by a curve. We represent it by the curve ci (α) : [0, 1] → R2 cover of an undirected graph Ḡ is a subset Vc of vertices of
such that ci (0.5) = xi with length p. We determine pairs V̄ such that if (vi , vj ) is an edge in Ḡ, then either vi ∈ V̄ or
of pixels which are mirror reflections of each other with a vj ∈ V̄. In order to find the minimum vertex cover, we solve
confidence score based on the following observation. the following integer linear program (ILP).
Let xi , xi0 ∈ F be any two edge pixels. If the pixels xi P
and xi0 are mirror reflections of each other and the image I min v∈V̄ xv
is mirror symmetric in the proximity of the pixels xi and xi0 , s.t. xu + xv ≥ 1 ∀(u, v) ∈ Ē
then the following equalities hold true. xv ∈ {0, 1} ∀v ∈ V̄ (4)
ci0 (α) = Rii0 QR>
ii0 (ci (α) − tii0 ) + tii0 , ∀α ∈ [0, 1] (1) Here, the binary variable xv is equal to 1, if the vertex v is in
ci (α) = Rii0 QR>
ii0 (ci0 (α) − tii0 ) + tii0 , ∀α ∈ [0, 1].(2) the vertex cover Vc and 0, otherwise. The constraint xu +xv ≥
1 ensures that at least one vertex of the edge (u, v) ∈ Ē is
Let η i (α) be the normal to the curve ci (α) at α. It is trivial
included in the vertex cover. We rewrite the above ILP in the
to show that if the curve ci and ci0 are mirror symmetric
standard form as below
then η ii0 (α) = η i0 (α) and η i0 i (α) = η i (α). Here, η i0 i (α) is
normal to the curve Rii0 QR> ii0 (ci0 (α) − tii0 ) + tii0 at α, and argmin 1> x subject to Ex ≥ 1, x ∈ {0, 1}|V̄| . (5)
η ii0 (α) is normal to the curve Rii0 QR> ii0 (ci (α) − tii0 ) + tii0
x
at α. Here, 1 is a vector of size |V̄| with all elements equal to 1 and
Therefore, if the pixels xi and xi0 are mirror reflections of the matrix E ∈ {0, 1}|Ē|×|V̄| is the edge incident matrix such
each other, then we have that that E(e, v) = 1 if the e-th edge is incident on the vertex v
Z 1 Z 1
and zero, if the e-th edge is not incident on the vertex v. The
(η >0
ii (α)η i (α) + η >
ii 0 (α)η i (α))dα = (1 + 1)dα = 2. vertex cover problem is an NP-hard problem. Therefore, we
0 0
(3) use the best known approximation which is 2-approximation
For the case of perfect symmetry, equation 3 holds true. obtained by relaxing the integer linear program in equation
However, due to the presence of noise and illumination varia- (5) to a linear program. In the relaxed program, each variable
tions, this might not hold true in practice. Furthermore, there takes value in [0, 1], i.e., x ∈ [0, 1]|V̄| . We obtain the final
may be various outlier pairs. Our goal is to detect all the solution by an optimal thresholding approach. If xi ≥ 0.5,
symmetry axes using the set of such pairs. First we cluster then xi = 1 and xi = 0, otherwise. Let Vc be the vertex
the pairs of mirror symmetric pixels. And then, we find the cover found. Then, the independent set I = V̄\Vc and the
symmetry axis in each cluster separately. Let {(xi , x0i )}vi=1 be clique C = I. We remove all the vertices of the clique C from
the detected pairs of mirror symmetric pixels. We construct the graph G and all edges incident on them. Then we find
an undirected graph G = (V, E), where each vertex vi in the the next dominant clique in the remaining graph. We find the
vertex set V corresponds to the pair (xi , xi0 ). We connect the first k dominant cliques by following the above procedure. We
vertices vi and vj by an unit weight edge if the symmetry present the complete procedure in Algorithm 1.
axes Li and Lj , defined by the pairs (xi , xi0 ) and (xj , xj 0 ) Using the pairs of mirror symmetric pixels detected by
respectively, are similar. We define the similarity between the Algorithm 1, we form the sets L and R by picking randomly
lines as follows. Let cj 0 i (α) = Rii0 QR>
ii0 (cj (α) − tii0 ) − tii0 one pixel of a pair and including it in the set L and the other
be the reflection of the curve through the symmetry axis pixel in the set R. We further remove any outlier pairs using
Li and η j 0 i (α) be its normal at α. Similarly, let ci0 j (α) = the following property of a symmetric function, since the pairs
4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
(a) (b)
3 90
1
120 60
0.8
2
150 0.6 30
1 0.4
0.2
0 180 0 0
-1
210 330
-2
-3 240 300
-6 -4 -2 0 2 4 6 270
(c) (d)
Fig. 2. Illustration of pairs of mirror symmetric pixels detection approach. (a) Three curves ci (α), cj (α), and ck (α). (b) Reflection of the curves ci (α),
cj (α) about the symmetry axes Lki and Lji defined by the pairs (ck , ci ) and (ck , cj ), respectively. (c)-(d) The normals η ki , η kj , η k to the curves cki ,
ckj , and ck , respectively, shown on the curve ck for better comparison.
Algorithm 1 Symmetry Detection randomization scheme is very high for h << |E|. Since the
1: symmetry present in the image is approximate symmetry, we
2: Input: Graph G, k=number of symmetry axes consider a pixel to be a mirror reflection even if it is shifted in a
3: for i ∈ {1, 2, . . . , k} do square of width u from its ideal location. Now, the probability
4: Construct complement graph Ḡ of graph G. of selecting the approximate mirror reflection pixel of a pixel
5: Find vertex cover Vc by solving 5 under consideration in one attempt is |E|−1 u2
. Therefore, the
6: Independent set I = V̄\Vc u 2
probability of not selecting in one attempt is 1 − |E|−1 .
7: Set cluster of MSPPs Pi = {(xj , xj 0 ) : j ∈ I}
Therefore, probability of not selecting the approximate mirror
8: Remove vertices I from the graph G and edges inci-
reflection pixelof a pixel under consideration in h attempts
dent on them. u2 h
9: end for is 1 − |E|−1 . Hence, the probability of selecting the
10: k sets of pairs of mirror symmetric points {Pi }ki=1 . u2
h reflection pixel of a pixel in h attempts
approximate mirror
is 1 − 1 − |E|−1 . For example, if |E| = 3000, h = 200, and
u = 5, this probability is 0.8124 which is quite high.
obtained are purely based on geometric constraints. Let the
points xi and xi0 be mirror reflections each other. Therefore,
I(xi ) = I(xi0 ) ⇒ ∇xi I(xi ) = ∇xi I(xi0 )
∇xi I(xi ) = ∇xi I(Rii0 QR> >
ii0 xi − Rii0 QRii0 tii0 + tii0 )
∇xi I(xi ) = Rii0 QR>
ii0 ∇xi0 I(xi0 ). (6)
(e)
Fig. 5. (a) Input image I, (b), detected edges, (c) the pairs of mirror symmetric points (L and R), (d) the initialized centers, the red centers are in the
non-symmetric regions and the green centers are in the symmetric regions, and (e) the symmetric superpixel segmentation with the pairs of mirror symmetric
superpixels.
0.85
0.14 0.92
Under Segmentation Error
0.8
Boundary Recall
0.12 0.9
0.75
0.1 0.88 SLIC
SEEDS
0.7 LSC
Proposed Approach
0.08 0.86
0.65
SLIC
0.06 SEEDS 0.84
0.6
LSC
Proposed Approach
0.04 0.55 0.82
100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000
Number of Superpixels Number of Superpixels Number of Superpixels
λ ∈ [1, 40]. We observe that using the proposed algorithm A. Symmetry Axis Detection
SymmSLIC, we are able to generate pairs of mirror symmetric We use the detected pairs mirror symmetric pixels in the
superpixels in the symmetric regions. In some cases, our Section III-A to detect the symmetry axes of the reflective
algorithm partitions a perceptually uniform region into many symmetric objects present in the input image. We represent
superpixels and due to the symmetric assignment strategy, the detected pairs of mirror symmetric pixels as the collection
we achieve a similar segmentation in the mirror symmetric of sets {Pi }ki=1 such that each set Pi contains pairs of mirror
8
proposed in [61]. In Fig. 8, we show the detected symmetry from [58]. We observe that our method does not require any
axes on few images from the dataset [58]. We compare our user interaction and still we get a comparable error rate. The
method using F-score with the methods [62], [63], [64], [65], performance of our method depends on how well the edges
[36], [35], for single symmetry axis detection on the dataset in are extracted in the given image.
[58]. In the TABLE I, we show the F -score for all the methods.
We observe that the proposed approach is able to achieve VI. C ONCLUSION
the state-of-the-art performance on the dataset in [58]. Here, In this work, we have proposed an algorithm to partition an
2tp
F -score = 2tp+f p+f n , tp = number of correctly detected image into superpixels, such that the symmetry present at the
axes, f p = number of incorrectly detected axes, and f n = pixel level is preserved at the superpixel level. We first detect
number missed ground-truth axes. the symmetry present at the pixel level, presented as pairs of
mirror symmetric pixels and then extend the SLIC algorithm
to preserve the symmetry at the superpixel level by proposing
a novel symmetric initialization and symmetric pixel center
assignment strategies. We observe that we are able to achieve
mirror symmetric superpixels in the symmetric regions. We
used detected pairs of mirror symmetric pixels to detect the
symmetry axes present in the image. We also proposed an
unsupervised symmetric objects segmentation approach using
the SymmSLIC superpixels and achieve accuracy close to
an unsupervised approach which requires human interface.
The main limitations of the proposed algorithm are that it
is applicable only to the fronto-parallel views containing
reflection symmetry and heavily depends on the performance
Fig. 8. Symmetry axes detection results on the dataset [58]. First two rows: of the edge detection algorithms. As a future work, we would
single symmetry axis, and last row: multiple symmetry axes. like to extend the proposed method for the rotation symmetry,
translation symmetry, and curved reflection symmetry. We also
would like to prepare a dataset containing the set of ground
B. Unsupervised Symmetric Object Segmentation truth pairs of superpixels for benchmarking the performance
of the algorithm developed.
Object segmentation is a challenging problem which is
generally performed either with user interaction and graph
ACKNOWLEDGEMENT
cuts ([66], [67], [68], [69]) or with supervised learning [70].
Our approach also differs from [71] in the sense that they Rajendra Nagar was supported by TCS Research Scholar-
perform local symmetry grouping whereas we perform global ship. Shanmuganathan Raman was supported by SERB-DST.
symmetry grouping. We would like to demonstrate how object
segmentation can be performed in an unsupervised manner
using SymmSLIC. This section is meant to illustrate as to R EFERENCES
how the algorithm developed in this paper can be used to [1] X. Ren and J. Malik, “Learning a classification model for segmentation.”
solve this classic computer vision application. The application in ICCV, vol. 1, 2003, pp. 10–17.
[2] J. Tighe and S. Lazebnik, “Superparsing: scalable nonparametric image
is however limited to images containing objects exhibiting parsing with superpixels,” Computer Vision–ECCV 2010, pp. 352–365,
reflection symmetry. 2010.
9
Fig. 9. Row 1: Input images, Row 2: the ground truth object segmentation, Row 3: grab cut in one cut [69] segmentation which requires a input mask around
the object, and Row 4: unsupervised symmetric objects segmentation using proposed approach which does not require any input mask.
[3] J. Wang, L. Ju, and X. Wang, “Image segmentation using local variation superpixels via geodesic distance,” International journal of computer
and edge-weighted centroidal voronoi tessellations,” IEEE Transactions vision, vol. 103, no. 1, pp. 1–21, 2013.
on Image Processing, vol. 20, no. 11, pp. 3242–3256, 2011. [17] R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. Süsstrunk,
[4] D. Hoiem, A. A. Efros, and M. Hebert, “Automatic photo pop-up,” ACM “Slic superpixels compared to state-of-the-art superpixel methods,” IEEE
transactions on graphics (TOG), vol. 24, no. 3, pp. 577–584, 2005. transactions on pattern analysis and machine intelligence, vol. 34,
[5] C. W. Tyler, Human symmetry perception and its computational analysis. no. 11, pp. 2274–2282, 2012.
Psychology Press, 2003. [18] R. Nagar and S. Raman, “Symmslic: Symmetry aware superpixel seg-
[6] Y. Sun and B. Bhanu, “Reflection symmetry-integrated image segmen- mentation,” in The IEEE International Conference on Computer Vision
tation,” IEEE transactions on pattern analysis and machine intelligence, (ICCV) Workshops, Oct 2017.
vol. 34, no. 9, pp. 1827–1841, 2012. [19] J. Shi and J. Malik, “Normalized cuts and image segmentation,” IEEE
[7] D. C. Hauagge and N. Snavely, “Image matching using local symmetry Transactions on pattern analysis and machine intelligence, vol. 22, no. 8,
features,” in Computer Vision and Pattern Recognition (CVPR), 2012 pp. 888–905, 2000.
IEEE Conference on. IEEE, 2012, pp. 206–213. [20] P. F. Felzenszwalb and D. P. Huttenlocher, “Efficient graph-based image
[8] S. Mitra and Y. Liu, “Local facial asymmetry for expression classifica- segmentation,” International journal of computer vision, vol. 59, no. 2,
tion,” in Computer Vision and Pattern Recognition, 2004. CVPR 2004. pp. 167–181, 2004.
Proceedings of the 2004 IEEE Computer Society Conference on, vol. 2. [21] Z. Li and J. Chen, “Superpixel segmentation using linear spectral
IEEE, 2004, pp. II–889. clustering,” in Proceedings of the IEEE Conference on Computer Vision
[9] G. Sela and M. D. Levine, “Real-time attention for robotic vision,” Real- and Pattern Recognition, 2015, pp. 1356–1363.
Time Imaging, vol. 3, no. 3, pp. 173–194, 1997. [22] O. Veksler, Y. Boykov, and P. Mehrani, “Superpixels and supervoxels in
[10] M. Mancas, B. Gosselin, and B. Macq, “Fast and automatic tumoral an energy optimization framework,” Computer Vision–ECCV 2010, pp.
area localisation using symmetry,” in Proceedings.(ICASSP’05). IEEE 211–224, 2010.
International Conference on Acoustics, Speech, and Signal Processing, [23] M. Van den Bergh, X. Boix, G. Roig, and L. Van Gool, “Seeds:
2005., vol. 2. IEEE, 2005, pp. 725–728. Superpixels extracted via energy-driven sampling,” International Journal
[11] K. Köser, C. Zach, and M. Pollefeys, “Dense 3d reconstruction of sym- of Computer Vision, vol. 111, no. 3, pp. 298–314, 2015.
metric scenes from a single image,” in Pattern Recognition. Springer, [24] Y. Zhang, R. Hartley, J. Mashford, and S. Burn, “Superpixels via
2011, pp. 266–275. pseudo-boolean optimization,” in Computer Vision (ICCV), 2011 IEEE
[12] S. N. Sinha, K. Ramnath, and R. Szeliski, “Detecting and reconstruct- International Conference on. IEEE, 2011, pp. 1387–1394.
ing 3d mirror symmetric objects,” in Computer Vision–ECCV 2012. [25] A. P. Moore, S. J. Prince, J. Warrell, U. Mohammed, and G. Jones,
Springer, 2012, pp. 586–600. “Superpixel lattices,” in Computer Vision and Pattern Recognition, 2008.
[13] Y. Ming, H. Li, and X. He, “Symmetry detection via contour grouping,” CVPR 2008. IEEE Conference on. IEEE, 2008, pp. 1–8.
in 2013 IEEE International Conference on Image Processing. IEEE, [26] L. Duan and F. Lafarge, “Image partitioning into convex polygons,” in
2013, pp. 4259–4263. Proceedings of the IEEE Conference on Computer Vision and Pattern
[14] Y.-J. Liu, C.-C. Yu, M.-J. Yu, and Y. He, “Manifold slic: A fast method Recognition, 2015, pp. 3119–3127.
to compute content-sensitive superpixels,” in Proceedings of the IEEE [27] A. Levinshtein, A. Stere, K. N. Kutulakos, D. J. Fleet, S. J. Dickinson,
Conference on Computer Vision and Pattern Recognition, 2016, pp. 651– and K. Siddiqi, “Turbopixels: Fast superpixels using geometric flows,”
659. IEEE transactions on pattern analysis and machine intelligence, vol. 31,
[15] M. Wertheimer, “Laws of organization in perceptual forms.” 1938. no. 12, pp. 2290–2297, 2009.
[16] P. Wang, G. Zeng, R. Gan, J. Wang, and H. Zha, “Structure-sensitive [28] M.-Y. Liu, O. Tuzel, S. Ramalingam, and R. Chellappa, “Entropy rate
10
Fig. 10. Results for the approaches TURBO [27], SLIC [17], ERS [28], LSC [21], SEEDS [23], MSLIC [14], and the proposed method. For each image, we
show two zoomed-in mirror symmetric windows to visualize whether the superpixels are mirror reflections of each other or not. Window border represents
its location in the image. The tilted windows are aligned horizontally.
11
superpixel segmentation,” in Computer Vision and Pattern Recognition [55] N. J. Mitra, M. Pauly, M. Wand, and D. Ceylan, “Symmetry in 3d
(CVPR), 2011 IEEE Conference on. IEEE, 2011, pp. 2097–2104. geometry: Extraction and applications,” in Computer Graphics Forum,
[29] Y. Liu, H. Hel-Or, C. S. Kaplan, and L. Van Gool, “Computational vol. 32, no. 6. Wiley Online Library, 2013, pp. 1–23.
symmetry in computer vision and computer graphics,” Foundations and [56] P. Arbelaez, M. Maire, C. Fowlkes, and J. Malik, “Contour detection and
Trends in Computer Graphics and Vision, vol. 5, no. 1-2, pp. 1–195, hierarchical image segmentation,” IEEE transactions on pattern analysis
2010. and machine intelligence, vol. 33, no. 5, pp. 898–916, 2011.
[30] A. Vasilier, “Recognition of symmetrical patterns in images,” in Inter- [57] J. Kleinberg and E. Tardos, Algorithm design. Pearson Education India,
national Conference on Pattern Recognition, 1984, pp. 1138–1140. 2006.
[31] J. L. Krahe, “Detection of symmetric and radial structures in images,” [58] C. Funk, S. Lee, M. R. Oswald, S. Tsogkas, W. Shen, A. Cohen,
in International Conference on Pattern Recognition, 1986, pp. 947–950. S. Dickinson, and Y. Liu, “2017 iccv challenge: Detecting symmetry
[32] S. K. Parui and D. D. Majumder, “Symmetry analysis by computer,” in the wild,” in The IEEE International Conference on Computer Vision
Pattern Recognition, vol. 16, no. 1, pp. 63–67, 1983. (ICCV) Workshops, Oct 2017.
[33] R. K. Yip, “A hough transform technique for the detection of reflectional [59] D. Martin, C. Fowlkes, D. Tal, and J. Malik, “A database of human
symmetry and skew-symmetry,” Pattern Recognition Letters, vol. 21, segmented natural images and its application to evaluating segmentation
no. 2, pp. 117–130, 2000. algorithms and measuring ecological statistics,” in Computer Vision,
[34] N. J. Mitra, L. J. Guibas, and M. Pauly, “Partial and approximate 2001. ICCV 2001. Proceedings. Eighth IEEE International Conference
symmetry detection for 3d geometry,” in ACM Transactions on Graphics on, vol. 2. IEEE, 2001, pp. 416–423.
(TOG), vol. 25, no. 3. ACM, 2006, pp. 560–568. [60] D. R. Martin, C. C. Fowlkes, and J. Malik, “Learning to detect natural
[35] G. Loy and J.-O. Eklundh, “Detecting symmetry and symmetric constel- image boundaries using local brightness, color, and texture cues,” IEEE
lations of features,” Computer Vision–ECCV 2006, pp. 508–521, 2006. transactions on pattern analysis and machine intelligence, vol. 26, no. 5,
[36] I. R. Atadjanov and S. Lee, “Reflection symmetry detection via ap- pp. 530–549, 2004.
pearance of structure descriptor,” in European Conference on Computer [61] R. Nagar and S. Raman, “Reflection symmetry axes detection using
Vision. Springer, 2016, pp. 3–18. multiple model fitting,” IEEE Signal Processing Letters, vol. 24, no. 10,
[37] E. Yodogawa, “Symmetropy, an entropy-like measure of visual symme- pp. 1438–1442, 2017.
try,” Perception & Psychophysics, vol. 32, no. 3, pp. 230–240, 1982. [62] E. Michaelsen and M. Arens, “Hierarchical grouping using gestalt
[38] G. Marola, “On the detection of the axes of symmetry of symmetric assessments,” in Proceedings, ICCV Workshop on Detecting Symmetry
and almost symmetric planar images,” IEEE Transactions on Pattern in the Wild, Venice, vol. 3, no. 4, 2017, p. 5.
Analysis and Machine Intelligence, vol. 11, no. 1, pp. 104–108, 1989. [63] M. Elawady, C. Ducottet, O. Alata, C. Barat, and P. Colantoni, “Wavelet-
[39] D. Shen, H. H. Ip, K. K. Cheung, and E. K. Teoh, “Symmetry detection based reflection symmetry detection via textural and color histograms:
by generalized complex (gc) moments: a close-form solution,” IEEE Algorithm and results,” in Computer Vision Workshop (ICCVW), 2017
Transactions on Pattern Analysis and Machine Intelligence, vol. 21, IEEE International Conference on. IEEE, 2017, pp. 1734–1738.
no. 5, pp. 466–476, 1999. [64] F. Guerrini, A. Gnutti, and R. Leonardi, “Innerspec: Technical report,”
[40] D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” in Proceedings, ICCV Workshop on Detecting Symmetry in the Wild,
International journal of computer vision, vol. 60, no. 2, pp. 91–110, Venice, vol. 3, 2017.
2004. [65] M. Cicconet, D. G. Hildebrand, and H. Elliott, “Finding mirror sym-
[41] S. Kondra, A. Petrosino, and S. Iodice, “Multi-scale kernel operators for metry via registration and optimal symmetric pairwise assignment of
reflection and rotation symmetry: further achievements,” in Proceedings curves: Algorithm and results,” in Computer Vision Workshop (ICCVW),
of the IEEE Conference on Computer Vision and Pattern Recognition 2017 IEEE International Conference on. IEEE, 2017, pp. 1759–1763.
Workshops, 2013, pp. 217–222. [66] C. Rother, V. Kolmogorov, and A. Blake, “Grabcut: Interactive fore-
[42] V. Patraucean, R. Grompone von Gioi, and M. Ovsjanikov, “Detection ground extraction using iterated graph cuts,” in ACM transactions on
of mirror-symmetric image patches,” in Proceedings of the IEEE Con- graphics (TOG), vol. 23, no. 3. ACM, 2004, pp. 309–314.
ference on Computer Vision and Pattern Recognition Workshops, 2013, [67] H. Fu, X. Cao, Z. Tu, and D. Lin, “Symmetry constraint for foreground
pp. 211–216. extraction,” IEEE transactions on cybernetics, vol. 44, no. 5, pp. 644–
[43] E. Michaelsen, D. Muench, and M. Arens, “Recognition of symmetry 654, 2014.
structure by use of gestalt algebra,” in Proceedings of the IEEE Con- [68] M. Tang, I. Ben Ayed, D. Marin, and Y. Boykov, “Secrets of grabcut and
ference on Computer Vision and Pattern Recognition Workshops, 2013, kernel k-means,” in Proceedings of the IEEE International Conference
pp. 206–210. on Computer Vision, 2015, pp. 1555–1563.
[44] S. Tsogkas and I. Kokkinos, “Learning-based symmetry detection in [69] M. Tang, L. Gorelick, O. Veksler, and Y. Boykov, “Grabcut in one cut,”
natural images,” in European Conference on Computer Vision. Springer, in Computer Vision (ICCV), 2013 IEEE International Conference on.
2012, pp. 41–54. IEEE, 2013, pp. 1769–1776.
[45] R. Scognamillo, G. Rhodes, C. Morrone, and D. Burr, “A feature-based [70] B. Hariharan, P. Arbeláez, R. Girshick, and J. Malik, “Simultaneous
model of symmetry detection,” Proceedings of the Royal Society of detection and segmentation,” in European Conference on Computer
London B: Biological Sciences, vol. 270, no. 1525, pp. 1727–1733, 2003. Vision. Springer, 2014, pp. 297–312.
[46] T. Zielke, M. Brauckmann, and W. Vonseelen, “Intensity and edge- [71] A. Levinshtein, C. Sminchisescu, and S. Dickinson, “Multiscale sym-
based symmetry detection with an application to car-following,” CVGIP: metric part detection and grouping,” International journal of computer
Image Understanding, vol. 58, no. 2, pp. 177–190, 1993. vision, vol. 104, no. 2, pp. 117–134, 2013.
[47] C. Sun, “Symmetry detection using gradient information,” Pattern
Recognition Letters, vol. 16, no. 9, pp. 987–996, 1995.
[48] M. Cho and K. M. Lee, “Bilateral symmetry detection via symmetry-
growing.” in BMVC. Citeseer, 2009, pp. 1–11.
[49] M. Bokeloh, A. Berner, M. Wand, H.-P. Seidel, and A. Schilling,
“Symmetry detection using feature lines,” in Computer Graphics Forum,
vol. 28, no. 2. Wiley Online Library, 2009, pp. 697–706.
[50] Q. Mo and B. Draper, “Detecting bilateral symmetry with feature
mirroring,” in CVPR Workshops, 2011.
[51] Z. Wang, Z. Tang, and X. Zhang, “Reflection symmetry detection using
locally affine invariant edge correspondence,” Image Processing, IEEE
Transactions on, vol. 24, no. 4, pp. 1297–1301, 2015.
[52] H. Cornelius and G. Loy, “Detecting bilateral symmetry in perspective,”
in 2006 Conference on Computer Vision and Pattern Recognition Work-
shop (CVPRW’06). IEEE, 2006, pp. 191–191.
[53] T. Tuytelaars, A. Turina, and L. Van Gool, “Noncombinatorial detection
of regular repetitions under perspective skew,” IEEE TPAMI, vol. 25,
no. 4, pp. 418–432, 2003.
[54] Z. Tang, P. Monasse, and J.-M. Morel, “Reflexive symmetry detection
in single image,” in International Conference on Curves and Surfaces.
Springer, 2014, pp. 452–460.