Photo Rating of Facial Pictures based on Image Segmentation

Arnaud Lienhard, Marion Reinhard, Alice Caplier and Patricia Ladret

GIPSA-lab, Grenoble University, 11, rue des Mathematiques, Grenoble, France

Keywords:

Aesthetic Image Quality, Portrait, Categorization, Image Segmentation.

Abstract:

A single glance at a face is enough to infer a ﬁrst impression about someone. With the increasing amount

of pictures available, selecting the most suitable picture for a given use is a difﬁcult task. This work focuses

on the estimation of the image quality of facial portraits. Some image quality features are extracted such

as blur, color representation, illumination and it is shown that concerning facial picture rating, it is better to

estimate each feature on the different picture parts (background and foreground). The performance of the

proposed image quality estimator is evaluated and compared with a subjective facial picture quality estimation

experiment.

1 INTRODUCTION

The development of digital cameras enable people to

have access to a constantly growing number of pho-

tos. Besides, selecting good-looking images is impor-

tant in a world where everybody is constantly look-

ing at others pictures. Social psychology studies show

that a 100 milliseconds exposure time to a facial por-

trait is sufﬁcient for people to appraise the subject at-

tractiveness, trustworthiness, competence or aggres-

siveness (Willis and Todorov, 2006).

The choice of a good facial picture may be ap-

plication dependent. Proﬁle pictures on social net-

works are different from pictures selected in a pro-

fessional purpose. Thus, the features used for auto-

matic aesthetic scoring have to be adapted to the con-

sidered application. Aesthetic rules depend on im-

age composition, and evaluating a landscape is differ-

ent from judging a portrait, where the viewer focuses

on the subject’s face. That is why ﬁnding faces and

their contours is important for good estimation per-

formance in portraiture aesthetics.

This work demonstrates the relevance of portraits

segmentation in the case of aesthetic scoring, by com-

paring the performance of a set of features computed

either on the entire image or separately on the seg-

mented image. Performance is measured and com-

pared with the results of a subjective quality estima-

tion experiment.

1.1 Related Work

Most of articles dealing with automatic photo rating

rely on machine learning techniques. Generally, the

whole process is the following. A dataset of pic-

tures rated by humans is created, in order to obtain

the ground truth. Then, a set of features is extracted

from these pictures. Features are related to photo-

graphic rules like the rule of thirds, explain global im-

age properties like the average luminance or describe

local image properties. Finally, a learning algorithm

is performed to create an image aesthetic model us-

ing the extracted features. An example of automatic

framework for aesthetic rating of photos is presented

in (Datta et al., 2006), and is publicly available on

http://acquine.alipr.com/.

Many feature sets have been proposed and tested.

(Ke et al., 2006) focus merely on high-level fea-

tures in order to separate poor quality from profes-

sional pictures. They design features corresponding

to semantic information and abstract concepts such

as composition, color or lighting. In their approach,

(Luo and Tang, 2008) include image composition fea-

tures by separating the image into two parts: a region

with sharp edges and a blurry region. The blurry re-

gion is deﬁned as the background, and features are

computed in both regions. However they did not focus

on portraits. Then, (Li and Gallagher, 2010) perform

face detection and include features related to human

faces and their positions: sizes, distances, etc. They

compute technical and perceptual features like con-

trast, blurring or colorfulness on both face and non

329

Lienhard A., Reinhard M., Caplier A. and Ladret P..

Photo Rating of Facial Pictures based on Image Segmentation.

DOI: 10.5220/0004673003290336

In Proceedings of the 9th International Conference on Computer Vision Theory and Applications (VISAPP-2014), pages 329-336

ISBN: 978-989-758-004-8

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

face regions. (Khan and Vogel, 2012) adapt spatial

composition features to portraiture classiﬁcation and

show the impact of the face position in aesthetic eval-

uation. However facial portraits require a more accu-

rate segmentation than only separating facial regions

from the background. For instance, hair or hats have

to be considered when evaluating the visual appeal of

a portrait.

1.2 Proposed Method

Our work focuses on frontal facial portraits, which

became very common with the proliferation of digital

cameras and social media websites.

The idea developed in this article is to show how

automatic image segmentation can improve the per-

formance of image quality evaluation. Images are

automatically segmented in 4 parts: facial skin, hair,

shoulders and background. This segmentation deﬁnes

two distinct areas. The foreground corresponds to the

subject’s face, which contains facial skin, hair, and

generally the top of the shoulders. The background

region is deﬁned as the part of the picture which is

not the foreground. This differs from previous work,

since (Li and Gallagher, 2010) and (Khan and Vogel,

2012) only performed face detection and did not ﬁnd

the precise contours of the facial regions. For estima-

tion, only few features are considered to emphasize

the importance of image segmentation. Considered

features are described in Section 3.2.

This paper is organized as follows. Section 2 de-

scribes the dataset that has been built and scored by

humans in order to have the ground truth. The quality

image features considered and the segmentation tech-

nique are explained in Section 3. Experimental results

of aesthetic quality evaluation are given in Section 4.

2 HUMAN RATING OF FACE

IMAGES

2.1 Database

A database containing 125 female and 125 male

digital pictures is considered. Half of the images

was extracted from free image datasets, such as La-

beled Faces in the Wild (Huang and Mattar, 2008)

and Caltech Face Dataset (see http://www.vision.

caltech.edu/html-ﬁles/archive.html). The other half

of facial images was acquired from private collection.

Most of them are consumer photos. They were

cropped to the extremes of the targets head and shoul-

ders (top of the head, bottom of the shoulders, sides of

hair or ears) and standardized for size (from 240×148

to 240 × 320).

The face in the picture is always right-side-up and

centered. A large variety of photos is considered: all

types of gaze and expression, facial hair styles, pres-

ence of accessories (earrings, eyeglasses, hat and vis-

ible make-up), clothing, etc.

2.2 Human Picture Rating

Twenty-ﬁve men and women, mostly aged from 20 to

30, were asked to judge facial photos using a discrete

scale from 1 (very low quality) to 6 (very high qual-

ity). They were instructed that they would be seeing

a series of faces randomly presented on a computer

screen and that their task is to assess the aesthetic

quality of each of the images displayed. Instructions

were the same for each participant: A picture will be

judged as very aesthetic when the numerous aspects

that qualify it (framing, luminosity, contrast, bloom-

ing, balance between the components, etc.) are of

good quality. Please indicate the aesthetic quality of

the picture, using a scale from 1 (very bad quality) to

6 (excellent quality).

Even if only facial portraits are considered, facial

beauty is not part of the criteria and the ﬁnal objective

is to evaluate the global aesthetic quality of the picture

only.

2.3 Preliminary Results

Each image is associated with a mean score over the

25 votes and a distribution of scores corresponding

to individual votes. These scores rank from 1.36 to

5.36 (Mean of 3.21, Standard Deviation of 0.73). The

distribution is largely Gaussian and all the possible

votes are represented, as presented in Figure 1. To

estimate the reliability of the test scores, Cronbach’s

Alpha (Cronbach, 1951) is computed: α = 0.99. α

ranges between 0 and 1 and it is commonly accepted

that α > 0.9 means a very high internal consistency

of the chosen scale.

These scores are considered as the ground truth

and are used to evaluate the performance of the image

quality estimator proposed in Section 4.

3 IMAGE SEGMENTATION AND

IMAGE QUALITY FEATURES

Finding the facial skin area may help a lot for aes-

thetic scoring in portraits. Computing blur or bright-

ness values in this special location may greatly differ

VISAPP2014-InternationalConferenceonComputerVisionTheoryandApplications

330

Figure 1: Histogram of image aesthetic quality mean rating.

from the global blur or brightness values, changing

our perception of the global image aesthetic.

3.1 Image Segmentation

3.1.1 Face and Facial Features Detection

Face bounding box detection is performed by using

the Viola-Jones algorithm described in (Viola and

Jones, 2001). It has been implemented in C++ us-

ing the OpenCV library, version 2.4.3. When several

faces are detected, only the biggest one is kept. De-

tection performance is good since all the faces among

the dataset of 250 images are detected.

The same algorithm is applied on the face bound-

ing box to detect eyes (left and right eyes are detected

separately), nose and mouth. Facial features detection

process has been tested on the LFW dataset (about

13,000 images) and obtained 89% of correct feature

extraction.

3.1.2 Facial Image Segmentation

The Viola-Jones algorithm provides bounding boxes

for face and facial features. Since our goal is to show

that a ﬁne segmentation of the face is required for a

good image quality estimation, the proposed method

for face segmentation is presented below. It is based

on the results obtained after applying the Viola-Jones

algorithm.

In order to detect the location of facial contours,

the advantages of two techniques are combined (see

Figure 3). First, a histogram containing the values of

skin pixels in the area deﬁned by the facial features

is created. After normalization, for each pixel value

in the entire image, the probability to be part of facial

skin is computed. Many skin detection models have

been developed, for example (Jones and Rehg, 1999)

used BGR color space and constructed histogram us-

ing hand labeled skin regions. Hue and saturation

channels from the HSV color space are often used.

However hue becomes less discriminant when dealing

with underexposed images or with different skin col-

ors. For that reason, red and saturation channels from

respectively BGR and HSV color spaces are used.

A segmentation algorithm is performed on the

probability image obtained from the histogram model,

to separate skin and non skin areas. Watershed is

widely used to solve segmentation problems. The

version implemented in OpenCV and inspired by

(Beucher and Meyer, 1993) is chosen. Skin area is

used as a seed for facial skin region, and image cor-

ners are used as seeds for the non skin area. Perform-

ing Watershed on the histogram back projection (see

Figure 3) prevents the algorithm from setting non skin

pixels to the facial skin area, and increases the initial

skin area deﬁned by facial feature detection.

Using this new skin area enables us to compute a

more precise histogram, which provides a better seed

for the next iteration of segmentation. Iterating this

several times often creates ﬁne contours. The whole

process is iterated 20 times.

To segment hair, we use exactly the same tech-

nique than for skin segmentation. The only difference

is the initialization step. Since we already have found

the skin area, and knowing that hair is above facial

skin, the area that is used to create the hair model is

deﬁned. Same operation is done for shoulder segmen-

tation: the initial shoulder location is the image part

below facial skin region.

Finally, the background region is deﬁned as the

remaining region, while the foreground region is the

addition of facial skin, hair and shoulders areas. Some

segmentation results are provided in Figure 2.

Figure 2: First row presents nice segmentation results,

while second row shows segmentation on more challenging

images: presence of sunglasses, unclear limitations between

skin, hair and background, etc.

3.2 Image Quality Features

6 measurements related to image quality are imple-

mented and are computed separately either in the

whole image or in subregions.

PhotoRatingofFacialPicturesbasedonImageSegmentation

331

Figure 3: Presentation of the segmentation process.

3.2.1 Blur

Blurry images are generally correlated to poor quality.

However this is not completely true with portraits: if

the face contains sharp edges, the background may be

blurred without affecting the global image aesthetic

quality. It may even enhance the visual appeal of the

picture, by increasing the contrast between face and

background.

The blur value is computed using the procedure

described in (Crete and Dolmiere, 2007). It is close

to 0 for a sharp image, and close to 1 for a blurred

image. The idea is to blur the initial image and to

compare the intensity variation of neighboring pixels

in both initial and blurred images. Figure 4 brieﬂy

describes the blur estimation principle. A couple of

images and their blur values are presented in Figure

Figure 4: Simpliﬁed ﬂowchart of the blur estimation princi-

ple described in (Crete and Dolmiere, 2007).

Figure 5: Examples of high and low blur levels, computed

in the entire image. Blur from left to right: F

= 0.29 and

= 0.62. Ground truth quality scores are 4.4 and 2.0, re-

spectively.

3.2.2 Color Count

The number of colors in an image may have an impact

on the global aesthetic feeling. Figure 6 shows exam-

ple of images with high and low colorfulness values.

Many articles use hue counts (Ke et al., 2006) or the

number of colors represented in the RGB color space

(Luo and Tang, 2008). We chose to use the Lab color

space, since it is designed to approximate human vi-

sion, and the color channels (a and b) are separated

from the luminance channel (L).

A 2-dimensional histogram H is computed from

both color channels, using 16 × 16 bins. We do not

take into account pixels with too low or too high lu-

minance values L (L < 40 or L > 240), since their

color values are not signiﬁcant. The histogram is nor-

malized in order to have a maximum value of 1. Let

H(i, j) be the frequency of pixels such that the a and

b color values belong to bins i and j, respectively.

We deﬁne |C| as the number of pairs (i, j) such that

H(i, j) > α. α is set to 0.001. The color count is ﬁ-

nally:

= |C|/256 × 100% (1)

equals to 1 when all the possible colors are repre-

sented, and is close to 0 otherwise.

Figure 6: Pictures with many different colors have a higher

colorfulness value than pictures with a uniform background.

From left to right: F

= 0.11 and F

= 0.02. Ground truth

quality scores are 3.4 and 2.0, respectively.

3.2.3 Illumination and Saturation

Professional photographers often adapt the brightness

of a picture, and make the face illumination differ-

VISAPP2014-InternationalConferenceonComputerVisionTheoryandApplications

332

ent from the background enlightening. The overall

brightness should neither be too high nor too low. In

this work, brightness average and standard deviation

are considered. Computing these feature values can

be done by considering the value channel V from the

HSV color space.

The same operation is done for the saturation

channel S. Saturation is a color purity indicator. Sat-

uration average and standard deviation are computed

from the saturation channel of the HSV color space.

After computation, the following values are ob-

tained:

• µ

, brightness mean

• σ

, brightness standard deviation

• µ

, saturation mean

• σ

, saturation standard deviation

3.2.4 Final Feature Set

The ﬁnal set of measurements F is

F = (F

,µ

,σ

,µ

,σ

)

This measurements are computed in the 6 following

regions:

1. Entire image

2. Foreground: addition of skin, hair and shoulders

3. Background

4. Facial skin area

5. Hair area

6. Shoulders area

This makes a total of 6 × 6 features. In the experi-

ments, the feature set F

is the set F computed on the

region i. For example, F

represents the set of fea-

tures F evaluated on the entire image.

4 AUTOMATIC AESTHETIC

QUALITY ESTIMATION

In this section, classiﬁcation and image scoring are

performed to show how much segmentation improves

the quality evaluation. Feature are tested separately

in section 4.2. In all the experiments, results are com-

pared with ground truth scores. 3 feature set combi-

nations are used to show segmentation inﬂuence:

1. Feature set F

is used to have an idea of the global

features performance, when they are computed on

the entire image.

2. Feature sets F

(foreground) and F

(background)

are used to demonstrate the segmentation inﬂu-

ence.

3. Finally, feature sets F

(background), F

(face),

(hair) and F

(shoulders) are used to show the

performance improvements when the foreground

is separated in 3 parts.

We used the k-fold cross validation technique,

with k = 10. Images were removed when we could

not extract all of the facial features (e.g. images with

large sunglasses).

4.1 Inﬂuence of Segmentation on

Classiﬁcation

The main objective of picture classiﬁcation is to sepa-

rate good looking images from bad pictures. Support

vector machines (SVM) are widely used to perform

categorization applied to image aesthetic evaluation

(Datta et al., 2007). To classify the pictures in several

groups, SVM with a Gaussian Kernel is performed.

Performance is evaluated by the Cross-Category Er-

ror (CCE)

CCE

∑

n=1

I( ˆc

− c

= i)

and the Multi-Category Error (MCE)

MCE =

−1

∑

i=−(N

−1)

|i|CCE(i)

where N

is the number of test images, N

the number

of classes, ˆc

the ground truth classiﬁcation described

in Section 2.3, c

the predicted classiﬁcation. i is the

difference between ground truth and predicted classi-

ﬁcation and I(.) is the indicator function.

Since random training and testing sets are used,

and due to the limited amount of images available

in the dataset, all of the experiments are repeated 10

times with various sets. Then, the average perfor-

mance is displayed.

4.1.1 2-Class Classiﬁcation

2-class categorization is used to separate high and low

quality images. Images below the median aesthetic

ground truth score are labeled as group 0, and im-

ages above as group 1. The median aesthetic score

is around 3.2 (see Figure 1). Results of the 3 exper-

iments described in Section 4 are presented in Table

It has to be noticed that results are signiﬁcantly above

the performance of a random classiﬁer. Results show

that segmenting the image can help a lot to classify

PhotoRatingofFacialPicturesbasedonImageSegmentation

333

Table 1: Experimental results for 2-class categorization.

CCE

−1

is the number of overrated images, CCE

is the

number of good classiﬁcation and CCE

is the number of

underrated images. Performance is the good classiﬁcation

rate.

F. sets Perf. CCE

−1

CCE

75.2% 22 128 20

, F

77.7% 23 132 15

...F

83.7% 20 141 9

images in two categories. After segmentation the

good classiﬁcation rate increases from 75% to almost

84%.

4.1.2 3-Class Classiﬁcation

3-Class categorization is an interesting problem, since

we often want to separate highly rated or remove

poorly rated images. For this task, 3 groups of about

50 images are used. The ﬁrst group contains the

50 lowest aesthetic score, the second group contains

images with medium aesthetic score (between 3 and

3.4). The last group contains only high-rated images.

We are using a total of 160 images. Images have label

0 for low aesthetic quality, 1 for medium quality and

2 for high quality.

The same experiments are conducted, and classi-

ﬁcation performance is presented in Table 2.

Table 2: Experimental results for 3-class categorization.

MCE measures the number of misclassiﬁcations.

F. sets Performance MCE

54.3% 85

, F

58% 80

...F

60.0% 78

A random classiﬁer would have a correct classiﬁca-

tion rate of about 33%. Again, segmentation helps for

3-Class categorization, since performance increases

from 54% to 60% after segmentation.

4.2 Inﬂuence of Image Quality Features

To have an idea of features inﬂuence, 2-Class cate-

gorization is performed using only one feature at a

time. Classiﬁcation performance is reported in Table

3 for the 6 features evaluated on the entire image (F

the foreground region (F

) and the background region

The ground truth quality score is highly correlated

with the blur value F

: there is about 70% of correct

classiﬁcation for the dataset considered. This score

change with respect to the image region considered.

The blur value is more discriminant when computed

on the foreground region (71%, and only 66% for the

Table 3: Features good classiﬁcation rates (%).

F. set F

70 52 55 59 54 51

71 50 57 52 57 51

66 44 49 55 50 51

background region), which is the most important part

of a portrait and should not be blurry. It is almost

the performance obtained by using the entire set F

(75%).

Other features like the color count F

or the sat-

uration standard deviation σ

have lower good clas-

siﬁcation rates, with respectively 52 and 51% when

computed on the entire image, which is not signiﬁ-

cantly higher than the 50% obtained using a random

classiﬁer.

In a portrait, a nice illumination for the subject

is important and face illumination has an impact on

image quality: the brightness mean µ

computed on

the foreground region has a good classiﬁcation rate of

57%. The same feature evaluated on the background

region has a lower score of 49%.

4.3 Inﬂuence of Segmentation on Image

Quality Scoring

In this section, the results of quality scoring with

and without image segmentation are presented. To

this end, a Support Vector Regression (SVR) is per-

formed. All of the images available in the dataset are

used: 90% for learning and 10% for testing. An SVR

implementation is given in the OpenCV library.

Measuring the regression performance is slightly

more difﬁcult. A model which always predicts the

average value will have a nice mean error if many im-

ages have aesthetic scores close to this mean. After

regression, the predicted data should be as close as

possible to the expected values.

To measure prediction performance, 3 criteria are

used. The Mean Squared Error E quantiﬁes the differ-

ence between ground truth and predicted scores. Let

ˆs

be the ground truth and s

the predicted score of

picture n:

E =

∑

n=1

( ˆs

− s

)

The Pearson correlation R is deﬁned as

R =

∑

n=1

( ˆs

−

ˆs) · (s

− ¯s)

∑

n=1

( ˆs

−

ˆs)

∑

n=1

− ¯s)

VISAPP2014-InternationalConferenceonComputerVisionTheoryandApplications

334

Figure 7: Prediction results for segmented and non segmented images.

where

ˆs =

∑

n=1

ˆs

and ¯s =

∑

n=1

. Finally the

Spearman rank correlation ρ is computed

ρ = 1 −

∑

n=1

− 1)

where d

= ˆs

−s

is the rank difference between vari-

ables. While Pearson’s measure quantiﬁes how much

the data are linearly correlated, Spearman’s rank cor-

relation tells if the data are monotonically correlated.

Values close to 1 result from correlated data and 0

means that the data are not correlated.

Figure 7 shows the prediction results after exper-

iments and Table 4 displays the 3 coefﬁcients com-

puted from the 3 experiments. An additional graph

is presented to show the result of a random computa-

tion. A perfect prediction would have only points on

the line y = x.

Table 4: Experimental results of regression.

Experiment Error E Corr. R Corr. ρ

0.59 0.52 0.53

, F

0.58 0.55 0.56

...F

0.55 0.61 0.64

Without any segmentation (Figure 7 b), the regres-

sion results are a lot better than randomized predic-

tion. However high and low scores may be far from

ground truth. Since the more common usages of aes-

thetic scoring are either eliminating low quality im-

ages or selecting high quality images, this is not a

satisfying result. Segmentations (Figures 7 c and d)

don’t reduce a lot the mean squared error (from 0.59

to 0.55, but reduce errors for extremes scores.

Image segmentation increases the correlation co-

efﬁcients values from about 0.5 to 0.6: the prediction

is more accurate. Examples of very high and very low

quality images are presented in Figure 8 with their

ground truth and predicted scores.

Figure 8: a) Highest ground truth scores. b) Lowest ground

truth scores. c) Highest predicted scores. d) Lowest pre-

dicted scores.

4.4 Validation on a Larger Dataset

In order to check the robustness of the method, we

have been looking for a larger dataset, containing

frontal facial images. Many photo sharing portals ex-

ist, like Photo.net or DPChallenge.com. These web-

sites allow people to share pictures and to score oth-

ers pictures. Datasets have been created by extract-

ing such images with their scores. We chose to ex-

tract a subset of the AVA database presented in (Mur-

ray, 2012), containing 250,000 images extracted from

DPChallenge. About 800 coloured frontal facial im-

ages were found, as described in Section 2.1.

To compare with the previous experiments, 2-

class categorization is performed using the 250 im-

ages with the lowest and highest aesthetic scores. Re-

sults are presented in Table 5.

Again, good classiﬁcation rate increases after seg-

mentation, from 55 to 63%. We still have interest-

ing results with simple features. However results are

worse than for the previous dataset. This may be ex-

plained by the use of more sophisticated photography

PhotoRatingofFacialPicturesbasedonImageSegmentation

335

Table 5: Experimental results for 2-class categorization of

the second dataset (500 images).

F. sets Perf. CCE

−1

CCE

55.3% 110 275 115

, F

58.6% 99 294 107

...F

63.0% 98 315 87

techniques in this dataset, making the features less ef-

fective. Moreover, it contains almost no blurry images

and performing 2-class categorization using only the

blur value computed on the foreground region leads to

an average performance of only 52.3% (71% for the

previous dataset). Pictures have a lower score vari-

ance (0.75 on a 1 to 10 scale) which makes the cate-

gorization difﬁcult.

5 CONCLUSIONS

In this work we proposed a method based on im-

age segmentation to explore aesthetics in portraits. A

few features were extracted, and image segmentation

techniques improved evaluation performance in both

categorization and score ranking tasks.

In the experiments, only a couple of hundred im-

ages have been used and it is not sufﬁcient to cre-

ate accurate models, especially for aesthetic scoring.

Gathering more images from different sharing portals

and other datasets may help a lot.

The features described and computed are simple

descriptors. They can be combined with generic im-

age descriptors, other data related to portraits, etc. Fa-

cial features like hair color, background composition

and textures, make-up and facial expressions, as well

as presence of hats and glasses can be used to provide

more accurate scoring.

Implementing new relevant features will be part

of future work. Comparison between several learning

techniques will be performed and additional regions

explored (e.g. eyes and mouth locations).

This will be a ﬁrst step in evaluating facial por-

traits with respect to other criteria like attractiveness,

competence, aggressiveness.

REFERENCES

Beucher, S. and Meyer, F. (1993). The morphological ap-

proach to segmentation: the watershed transforma-

tion. Mathematical Morphology in Image Processing,

pages 433–481.

Crete, F. and Dolmiere, T. (2007). The blur effect: percep-

tion and estimation with a new no-reference percep-

tual blur metric. Proc. of the SPIE, 6492.

Cronbach, L. (1951). Coefﬁcient alpha and the internal

structure of tests. Psychometrika, 16(3).

Datta, R., Joshi, D., Li, J., and Wang, J. Z. (2006). Studying

Aesthetics in Photographic Images Using a Computa-

tional Approach. ECCV, pages 288–301.

Datta, R., Li, J., and Wang, J. Z. (2007). Learning the

consensus on visual quality for next-generation image

management. Proc. of the 15th international confer-

ence on Multimedia, pages 533–536.

Huang, G. and Mattar, M. (2008). Labeled faces in the wild:

A database for studying face recognition in uncon-

strained environments. Workshop on Faces in ’Real-

Life’ Images: Detection, Alignment, and Recognition,

pages 1–11.

Jones, M. and Rehg, J. (1999). Statistical color models with

application to skin detection. CVPR, 1:274–280.

Ke, Y., Tang, X., and Jing, F. (2006). The design of high-

level features for photo quality assessment. CVPR,

1:419–426.

Khan, S. and Vogel, D. (2012). Evaluating visual aesthetics

in photographic portraiture. Proc.of the Eighth Annual

Symposium on Computational Aesthetics in Graphics,

Visualization, and Imaging (CAe ’12), pages 1–8.

Li, C. and Gallagher, A. (2010). Aesthetic quality assess-

ment of consumer photos with faces. ICIP, pages

3221 – 3224.

Luo, Y. and Tang, X. (2008). Photo and video quality eval-

uation: Focusing on the subject. ECCV, pages 386–

399.

Murray, N. (2012). AVA: A large-scale database for aes-

thetic visual analysis. CVPR, 0:2408–2415.

Viola, P. and Jones, M. (2001). Rapid object detection using

a boosted cascade of simple features. CVPR, 1:511–

518.

Willis, J. and Todorov, A. (2006). Making Up Your Mind

After a 100-Ms Exposure to a Face. Psychological

Science, 17(7):592–598.

VISAPP2014-InternationalConferenceonComputerVisionTheoryandApplications

336