Studying Aesthetics in Photographic Images Usinga Computational ...

Text-only Preview

Studying Aesthetics in Photographic Images
Using a Computational Approach
Ritendra Datta
Dhiraj Joshi
Jia Li
James Z. Wang
The Pennsylvania State University, University Park, PA 16802, USA
Abstract. Aesthetics, in the world of art and photography, refers to
the principles of the nature and appreciation of beauty. Judging beauty
and other aesthetic qualities of photographs is a highly subjective task.
Hence, there is no unanimously agreed standard for measuring aesthetic
value. In spite of the lack of firm rules, certain features in photographic
images are believed, by many, to please humans more than certain
others. In this paper, we treat the challenge of automatically inferring
aesthetic quality of pictures using their visual content as a machine
learning problem, with a peer-rated online photo sharing Website as
data source. We extract certain visual features based on the intuition
that they can discriminate between aesthetically pleasing and displeasing
images. Automated classifiers are built using support vector machines
and classification trees. Linear regression on polynomial terms of the
features is also applied to infer numerical aesthetics ratings. The work
attempts to explore the relationship between emotions which pictures
arouse in people, and their low-level content. Potential applications
include content-based image retrieval and digital photography.
Photography is defined as the art or practice of taking and processing pho-
tographs. Aesthetics in photography is how people usually characterize beauty
in this form of art. There are various ways in which aesthetics is defined by
different people. There exists no single consensus on what it exactly pertains to.
The broad idea is that photographic images that are pleasing to the eyes are
considered to be higher in terms of their aesthetic beauty. While the average
individual may simply be interested in how soothing a picture is to the eyes, a
photographic artist may be looking at the composition of the picture, the use
of colors and light, and any additional meanings conveyed by the picture. A
professional photographer, on the other hand, may be wondering how difficult
it may have been to take or to process a particular shot, the sharpness and the
color contrast of the picture, or whether the “rules of thumb” in photography
have been maintained. All these issues make the measurement of aesthetics in
pictures or photographs extremely subjective.
This work is supported in part by the US National Science Foundation, the
PNC Foundation, and SUN Microsystems. Corresponding author: R. Datta,
[email protected] More information:

In spite of the ambiguous definition of aesthetics, we show in this paper
that there exist certain visual properties which make photographs, in general,
more aesthetically beautiful. We tackle the problem computationally and exper-
imentally through a statistical learning approach. This allows us to reduce the
influence of exceptions and to identify certain features which are statistically
significant in good quality photographs.
Content analysis in photographic images has been studied by the multime-
dia and vision research community in the past decade. Today, several efficient
region-based image retrieval engines are in use [13, 6, 21, 18]. Statistical modeling
approaches have been proposed for automatic image annotation [4, 12]. Cultur-
ally significant pictures are being archived in digital libraries [7]. Online photo
sharing communities are becoming more and more common [1, 3, 11, 15]. In this
age of digital picture explosion, it is critical to continuously develop intelligent
systems for automatic image content analysis.
Community-based Photo Ratings as Data Source
One good data source is a large online photo sharing community,,
possibly the first of its kind, started in 1997 by Philip Greenspun, then a re-
searcher on online communities at MIT [15]. Primarily intended for photogra-
phy enthusiasts, the Website attracts more than 400, 000 registered members.
Many amateur and professional photographers visit the site frequently, share
photos, and rate and comment on photos taken by peers. There are more than
one million photographs uploaded by these users for perusal by the community.
Of interest to us is the fact that many of these photographs are peer-rated in
terms of two qualities, namely aesthetics and originality. The scores are given
in the range of one to seven, with a higher number indicating better rating.
This site acts as the main source of data
Plot of Aesthetics v/s Originality over 3581 photographs
for our computational aesthetics work. The
reason we chose such an online community
is that it provides photos which are rated
by a relatively diverse group. This ensures
generality in the ratings, averaged out over
the entire spectrum of amateurs to serious
professionals. While amateurs represent the
general population, the professionals tend to
spend more time on the technical details be-
Fig. 1. Correlation between the
fore rating the photographs. One caveat: The
aesthetics and originality ratings
nature of any peer-rated community is such
for 3581 photographs.
that it leads to unfair judgments under cer-
tain circumstances, and is no ex-
ception, making our acquired data fairly noisy. Ideally, the data should have
been collected from a random sample of human subjects under controlled setup,
but resource constraints prevented us from doing so.
We downloaded those pictures and their associated metadata which were
rated by at least two members of the community. For each image downloaded, we

parsed the pages and gathered the following information: (1) average aesthetics
score between 1.0 and 7.0, (2) average originality score between 1.0 and 7.0, (3)
number of times viewed by members, and (4) number of peer ratings.
Aesthetics and Originality
According to the Oxford Advanced Learner’s Dictionary, Aesthetics means (1)
“concerned with beauty and art and the understanding of beautiful things”, and
(2) “made in an artistic way and beautiful to look at”. A more specific discussion
on the definition of aesthetics can be found in [16]. As can be observed, no con-
sensus was reached on the topic among the users, many of whom are professional
photographers. Originality has a more specific definition of being something that
is unique and rarely observed. The originality score given to some photographs
can also be hard to interpret, because what seems original to some viewers may
not be so for others. Depending on the experiences of the viewers, the originality
scores for the same photo can vary considerably. Thus the originality score is
subjective to a large extent as well.
One of the first observations made on the
gathered data was the strong correlation be-
tween the aesthetics and originality ratings for
a given image. A plot of 3581 unique photo-
graph ratings can be seen in Fig. 1. As can be
seen, aesthetics and originality ratings have ap-
proximately linear correlation with each other.
This can be due to a number of factors. Many
users quickly rate a batch of photos in a given
Fig. 2. Aesthetics scores can
day. They tend not to spend too much time try-
be significantly influenced by
ing to distinguish between these two parameters
the semantics. Loneliness is
when judging a photo. They more often than
depicted using a person in this
not rate photographs based on a general impres-
frame, though the area occu-
sion. Typically, a very original concept leads to
pied by the person is very
good aesthetic value, while beauty can often be
small. Avg. aesthetics: 6.0/7.0
characterized by originality in view angle, color,
lighting, or composition. Also, because the rat-
ings are averages over a number of people, disparity by individuals may not be
reflected as high in the averages. Hence there is generally not much disparity in
the average ratings. In fact, out of the 3581 randomly chosen photos, only about
1.1% have a disparity of more than 1.0 between average aesthetics and average
originality, with a peak of 2.0.
As a result of this observation, we chose to limit the rest of our study to
aesthetics ratings only, since the value of one can be approximated to the value
of the other, and among the two, aesthetics has a rough definition that in prin-
ciple depends somewhat less on the content or the semantics of the photograph,
something that is very hard for present day machine intelligence to interpret ac-
curately. Nonetheless, the strong dependence on originality ratings means that
aesthetics ratings are also largely influenced by the semantics. As a result, some

visually similar photographs are rated very differently. For example in Fig. 2,
loneliness is depicted using a man in the frame, increasing its appeal, while the
lack of the person makes the photograph uninteresting and is likely to cause
poorer ratings from peers. This makes the task of automatically determining
aesthetics of photographs highly challenging.
Our Computational Aesthetics Approach
A classic treatise on psychological theories for understanding human perception
can be found in [2]. Here, we take the first step in using a computational approach
to understand what aspects of a photograph appeal to people, from a population
and statistical standpoint. For this purpose, we aim to build (1) a classifier that
can qualitatively distinguish between pictures of high and low aesthetic value,
or (2) a regression model that can quantitatively predict the aesthetics score,
both approaches relying on low-level visual features only. We define high or low
in terms of predefined ranges of aesthetics scores.
There are reasons to believe that classification may be a more appropriate
model than regression in tackling this problem. For one, the measures are highly
subjective, and there are no agreed standards for rating. This may render abso-
lute scores less meaningful. Again, ratings above or below certain thresholds on
an average by a set of unique users generally reflect on the photograph’s quality.
This way we also get around the problem of consistency where two identical
photographs can be scored differently by different groups of people. However, it
is more likely that both the group averages are within the same range and hence
are treated fairly when posed as a classification problem.
On the other hand, the ‘ideal’ case is when a machine can replicate the task
of robustly giving images aesthetics scores in the range of (1.0-7.0) the humans
do. This is the regression formulation of the problem. The possible benefits of
building a computational aesthetics model can be summarized as follows: If the
low-level image features alone can tell what range aesthetics ratings the image
deserves, this can potentially be used by photographers to get a rough estimate
of their shot composition quality, leading to adjustment in camera parameters or
shot positioning for improved aesthetics. Camera manufacturers can incorporate
a ‘suggested composition’ feature into their products. Alternatively, a content-
based image retrieval (CBIR) system can use the aesthetics score to discriminate
between visually similar images, giving greater priority to more pleasing query
results. Biologically speaking, a reasonable solution to this problem may lead to
a better understanding of the human vision.
Visual Feature Extraction
Experiences with photography lead us to believe in certain aspects as being
critical to quality. This entire study is on such beliefs or hypotheses and their
validation through numerical results. We treat each downloaded image separately
and extract features from them. We use the following notation: The RGB data of
each image is converted to HSV color space, producing two-dimensional matrices
IH , IS , and IV , each of dimension X × Y .

Our motivation for the choice of features was principled, based on (1) rules
of thumb in photography, (2) common intuition, and (3) observed trends in
ratings. In photography and color psychology, color tones and saturation play
important roles, and hence working in the HSV color space makes computation
more convenient. For some features we extract information from objects within
the photographs. An approximate way to find objects within images is segmen-
tation, under the assumption that homogeneous regions correspond to objects.
We use a fast segmentation method based on clustering. For this purpose the
image is transformed into the LU V space, since in this space locally Euclidean
distances model the perceived color change well. Using a fixed threshold for all
the photographs, we use the K-Center algorithm to compute cluster centroids,
treating the image pixels as a bag of vectors in LU V space. With these centroids
as seeds, a K-means algorithm computes clusters. Following a connected compo-
nent analysis, color-based segments are obtained. The 5 largest segments formed
are retained and denoted as {s1, ..., s5}. These clusters are used to compute
region-based features as we shall discuss in Sec. 2.7.
We extracted 56 visual features for each image. The feature set was care-
fully chosen but limited because our goal was mainly to study the trends or
patterns, if any, that lead to higher or lower aesthetics ratings. If the goal was
to only build a strong classifier or regression model, it would have made sense
to generate exhaustive features and apply typical machine-learning techniques
such as boosting. Without meaningful features it is difficult to make meaningful
conclusions from the results. We refer to our features as candidate features and
denote them as F = {fi|1 ≤ i ≤ 56} which are described as follows.
Exposure of Light and Colorfulness
Measuring the brightness using a light meter and a gray card, controlling the
exposure using the aperture and shutter speed settings, and darkroom print-
ing with dodging and burning are basic skills for any professional photographer.
Too much exposure (leading to brighter shots) often yields lower quality pictures.
Those that are too dark are often also not appealing. Thus light exposure can
often be a good discriminant between high and low quality photographs. Note
that there are always exceptions to any ‘rules of thumb’. An over-exposed or
under-exposed photograph under certain scenarios may yield very original and
beautiful shots. Ideally, the use of light should be characterized as normal day-
light, shooting into the sun, backlighting, shadow, night etc. We use the average
Y −1
pixel intensity f1 = 1
y=0 IV (x, y) to characterize the use of light.
We propose a fast and robust method to compute relative color distribution,
distinguishing multi-colored images from monochromatic, sepia or simply low
contrast images. We use the Earth Mover’s Distance (EMD) [17], which is a
measure of similarity between any two weighted distributions. We divide the
RGB color space into 64 cubic blocks with four equal partitions along each
dimension, taking each such cube as a sample point. Distribution D1 is generated
as the color distribution of a hypothetical image such that for each of 64 sample
points, the frequency is 1/64. Distribution D2 is computed from the given image

by finding the frequency of occurrence of color within each of the 64 cubes. The
EMD measure requires that the pairwise distance between sampling points in
the two distributions be supplied. Since the sampling points in both of them are
identical, we compute the pairwise Euclidean distances between the geometric
centers ci of each cube i, after conversion to LU V space. Thus the colorfulness
measure f2 is computed as follows: f2 = emd(D1, D2, {d(a, b) | 0 ≤ a, b ≤ 63}),
where d(a, b) = ||rgb2luv(ca) − rgb2luv(cb)|| .
Fig. 3. The proposed colorfulness measure. The two photographs on the left have high
values while the two on the right have low values.
The distribution D1 can be interpreted as the ideal color distribution of a
‘colorful’ image. How similar the color distribution of an arbitrary image is to
this one is a rough measure of how colorful that image is. Examples of images
producing high and low values of f2 are shown in Fig. 3.
Saturation and Hue
Saturation indicates chromatic purity. Pure colors in a photo tend to be more
appealing than dull or impure ones. In natural out-door landscape photography,
professionals use specialized film such as the Fuji Velvia to enhance the sat-
uration to result in deeper blue sky, greener grass, more vivid flowers, etc. We
Y −1
compute the average saturation f3 = 1
y=0 IS(x, y) as the saturation
indicator. Hue is similarly computed averaged over IH to get feature f4, though
the interpretation of such a feature is not as clear as the former. This is because
hue as defined in the HSV space corresponds to angles in a color wheel.
The Rule of Thirds
A very popular rule of thumb in photography is the Rule of Thirds. The rule can
be considered as a sloppy approximation to the ‘golden ratio’ (about 0.618). It
specifies that the main element, or the center of interest, in a photograph should
lie at one of the four intersections as shown in Fig. 4 (a). We observed that most
professional photographs that follow this rule have the main object stretch from
an intersection up to the center of the image. Also noticed was the fact that
centers of interest, e.g., the eye of a man, were often placed aligned to one of the
edges, on the inside. This implies that a large part of the main object often lies
on the periphery or inside of the inner rectangle. Based on these observations,
we computed the average hue as f5 = 9
y=Y/3 IH (x, y), with f6 and
f7 being similarly computed for IS and IV respectively.

Fig. 4. (a) The rule of thirds in photography: Imaginary lines cut the image horizontally
and vertically each into three parts. Intersection points are chosen to place important
parts of the composition instead of the center. (b)-(d) Daubechies wavelet transform.
Left: Original image. Middle: Three-level transform, levels separated by borders. Right:
Arrangement of three bands LH, HL and HH of the coefficients.
Familiarity Measure
We humans learn to rate the aesthetics of pictures from the experience gathered
by seeing other pictures. Our opinions are often governed by what we have
seen in the past. Because of our curiosity, when we see something unusual or
rare we perceive it in a way different from what we get to see on a regular
basis. In order to capture this factor in human judgment of photography, we
define a new measure of familiarity based on the integrated region matching
(IRM) image distance [21]. The IRM distance computes image similarity by using
color, texture and shape information from automatically segmented regions, and
performing a robust region-based matching with other images. Primarily meant
for image retrieval applications, we use it here to quantify familiarity. Given
a pre-determined anchor database of images with a well-spread distribution of
aesthetics scores, we retrieve the top K closest matches in it with the candidate
image as query. Denoting IRM distances of the top matches for each image
in decreasing order of rank as {q(i)|1 ≤ i ≤ K}. We compute f8 and f9 as
f8 = 1
i=1 q(i) , f9 = 1
i=1 q(i).
In effect, these measures should yield higher values for uncommon images.
Two different scales of 20 and 100 top matches are used since they may poten-
tially tell different stories about the uniqueness of the picture. While the former
measures average similarity in a local neighborhood, the latter does so on a more
global basis. Because of the strong correlation between aesthetics and originality,
it is intuitive that a higher value of f8 or f9 corresponds to greater originality
and hence we expect greater aesthetics score.
Wavelet-based Texture
Graininess or smoothness in a photograph can be interpreted in different ways.
If as a whole it is grainy, one possibility is that the picture was taken with a
grainy film or under high ISO settings. If as a whole it is smooth, the picture can
be out-of-focus, in which case it is in general not pleasing to the eye. Graininess
can also indicate the presence/absence and nature of texture within the image.
The use of texture is a composition skill in photography. One way to mea-
sure spatial smoothness in the image is to use Daubechies wavelet transform [10],

which has often been used in the literature to characterize texture. We perform
a three-level wavelet transform on all three color bands IH , IS and IV . An ex-
ample of such a transform on the intensity band is shown in Fig. 4 (b)-(c). The
three levels of wavelet bands are arranged from top left to bottom right in the
transformed image, and the four coefficients per level, LL, LH, HL, and HH are
arranged as shown in Fig. 4 (d). Denoting the coefficients (except LL) in level
i for the wavelet transform on hue image IH as whh
i , whl
and wlh
i , i = {1, 2, 3},
we define features f10, f11 and f12 as follows:
fi+9 =
i (x, y) +
i (x, y) +
i (x, y)
where Sk = |whh
i | + |whl
i | + |whh
i | and i = 1, 2, 3. The corresponding wavelet
features for saturation (IS ) and intensity (IV ) images are computed similarly
to get f13 through f15 and f16 through f18 respectively. Three more wavelet
features are derived. The sum of the average wavelet coefficients over all three
frequency levels for each of H, S and V are taken to form three additional
features: f19 =
i=10 fi, f20 =
i=13 fi, and f21 =
i=16 fi.
Size and Aspect Ratio
The size of an image has a good chance of affecting the photo ratings. Although
scaling is possible in digital and print media, the size presented initially must
be agreeable to the content of the photograph. A more crucial parameter is
the aspect ratio. It is well-known that 4 : 3 and 16 : 9 aspect ratios, which
approximate the ‘golden ratio,’ are chosen as standards for television screens or
70mm movies, for reasons related to viewing pleasure. The 35mm film used by
most photographers has a ratio of 3 : 2 while larger formats include ratios like
7 : 6 and 5 : 4. While size feature is f22 = X + Y , the aspect ratio feature is
f23 = X
Y .
Region Composition
Segmentation results in rough grouping of similar pixels,
which often correspond to objects in the scene. We denote
the set of pixels in the largest five connected components
or patches formed by the segmentation process described
before as {s1, ...s5}. The number of patches t ≤ 5 which
satisfy |si| ≥ XY
100 denotes feature f24. The number of color-
based clusters formed by K-Means in the LU V space is
Fig. 5. The HSV
feature f
Color Wheel.
25. This number is image dependent and dynami-
cally chosen, based on the complexity of the image. These
two features combine to measure how many distinct color blobs and how many
disconnected significantly large regions are present.
We then compute the average H, S and V values for each of the top 5 patches
as features f26 through f30, f31 through f35 and f36 through f40 respectively.
Features f41 through f45 store the relative size of each segment with respect to
the image, and are computed as fi+40 = |si|/(XY ) where i = 1, ..., 5.

The hue component of HSV is such that the colors that are 180◦ apart in
the color circle (Fig. 5) are complimentary to each other, which means that they
add up to ‘white’ color. These colors tend to look pleasing together. Based on
this idea, we define two new features, f46 and f47 in the following manner, cor-
responding to average color spread around the wheel and average complimentary
colors among the top 5 patch hues. These features are defined as
f46 =
|hi − hj|, f47 =
l(|hi − hj|), hi =
IH (x, y),
i=1 j=1
i=1 j=1
where l(k) = k if k ≤ 180◦, 360◦ − k if k > 180◦ . Finally, the rough po-
sitions of each segment are stored as features f48 through f52. We divide the
image into 3 equal parts along horizontal and vertical directions, locate the
block containing the centroid of each patch si, and set f47+i = (10r + c) where
(r, c) ∈ {(1, 1), ..., (3, 3)} indicates the corresponding block starting with top-left.
Low Depth of Field Indicators
Pictures with a simplistic composition and a well-focused center of interest are
sometimes more pleasing than pictures with many different objects. Professional
photographers often reduce the depth of field (DOF) for shooting single objects
by using larger aperture settings, macro lenses, or telephoto lenses. DOF is the
range of distance from a camera that is acceptably sharp in the photograph. On
the photo, areas in the DOF are noticeably sharper.
We noticed that a large number of low DOF photographs, e.g., insects, other
small creatures, animals in motion, were given high ratings. One reason may
be that these shots are difficult to take, since it is hard to focus steadily on
small and/or fast moving objects like insects and birds. A common feature is
that they are taken either by macro or by telephoto lenses. We propose a novel
method to detect low DOF and macro images. We divide the image into 16
equal rectangular blocks {M1, ...M16}, numbered in row-major order. Let w3 =
3 , whl
3 , whh
3 } denote the set of wavelet coefficients in the high-frequency (level
3 by the notation in Sec. 2.5) of the hue image IH . The low depth of field indicator
feature f53 for hue is computed as follows, with f54 and f55 being computed
similarly for IS and IV respectively:
w3(x, y)
53 =
(x,y)∈M w
3(x, y)
The object of interest in a macro shot is usually in sharp focus near the
center, while the surrounding is usually out of focus. This essentially means that
a large value of the low DOF indicator features tend to occur for macro shots.
Shape Convexity
It is believed that shapes in a picture also influence the degree of aesthetic
beauty perceived by humans. The challenge in designing a shape feature lies in
the understanding of what kind of shape pleases humans, and whether any such

Fig. 6. Demonstrating the shape convexity feature. Left: Original photograph. Middle:
Three largest non-background segments shown in original color. Right: Exclusive re-
gions of the convex hull generated for each segment are shown in white. The proportion
of white regions determine the convexity value.
measure generalizes well enough or not. As always, we hypothesize that convex
shapes like perfect moon, well-shaped fruits, boxes, or windows have an appeal,
positive or negative, which is different from concave or highly irregular shapes.
Let the image be segmented, as described before, and R patches {p1, ..., pR} are
obtained such that |pk| ≥ XY
200 . For each pk, we compute its convex hull, denoted
by g(pk). For a perfectly convex shape, pk ∩g(pk) = pk, i.e. |pk|
|g(pk)| = 1. We define
the shape convexity feature as f
k |
56 =
XY {
k=1 I( |g(pk)| ≥ 0.8)|pk|}, allowing
some room for irregularities of edge and error due to digitization. Here I(·) is
the indicator function. This feature can be interpreted as the fraction of the
image covered by approximately convex-shaped homogeneous regions, ignoring
the insignificant image regions. This feature is demonstrated in Fig. 6. Note that
a critical factor here is the segmentation process, since we are characterizing
shape by segments. Often, a perfectly convex object is split into concave or
irregular parts, considerably reducing the reliability of this measure.
Feature Selection, Classification, and Regression
A contribution of our work is the feature extraction process itself, since each fea-
ture represents an interesting aspects of photography. We now perform selection
in order to (1) discover features that show correlation with community-based
aesthetics scores, and (2) build a classification/regression model using a sub-
set of strongly/weakly relevant features such that generalization performance is
near optimal. Instead of using any regression model, we use a one-dimensional
support vector machine (SVM) [20]. SVMs are essentially powerful binary clas-
sifiers that project the data space into higher dimensions where the two classes
of points are linearly separable. Naturally, for one-dimensional data, they can
be more flexible than a single threshold classifier.
For the 3581 images downloaded, all 56 features in F were extracted and
normalized to the [0, 1] range to form the experimental data. Two classes of
data are chosen, high containing samples with aesthetics scores greater than
5.8, and low with scores less than 4.2. Only images that were rated by at least
two unique members were used. The reason for choosing classes with a gap is
that pictures with close lying aesthetic scores, e.g., 5.0 and 5.1 are not likely
to have any distinguishing feature, and may merely be representing the noise