IEEE International Conference on Computer, Communication and Signal Processing (ICCCSP-2017)
Implementation Of Virtual Fitting Room Using Image
Processing
Srinivasan K.
Department of Electronics and Instrumentation Engineering
Sri Ramakrishna Engineering College
Coimbatore, India
hod-eie@srec.ac.in
Abstract—There has been a great increase in interests towards
online shopping. In case of purchase of products like apparels
which always require a sense of knowledge on how cloths would
fit upon a person. This is the major reason why less number of
apparels are being shopped online. Hence, a virtual dressing
room which would make people know how cloths personally fits
in would be a great luxury for the online sellers which could give
a wide choice for customers. For online marketers, this would be
a great tool for enhancing its market.
Keywords—online shopping,color clustering,human
contour,virtual fitting.
I. INTRODUCTION
In recent years, particularly the last five years there has been
a hike in usage of internet in all the categories. The number of
people access the internet and utilizing internet for shopping
[4] keeps on increasing due to the development in the field of
information technology. Online marketing helps the producers
to bring out their varieties of products to a mass in the easiest
way. For customers, online shopping would give more
information and availability of all kinds of products in every
stream. This makes every product to come to the doorstep and
gives consumers the choice of taste and purchase. But when
this comes to dressing the quantity purchased is comparatively
less. This is because of the fact that people wish to know how
cloths looks on oneself and how both the top and bottom
matches together and also how the size of clothes fits the
contour of oneself. A Virtual Dressing Room is to fulfill all
the above necessities and would give the comfortability of
online shopping of clothes and hence a wide choice to
consumers.This also strengthens the platform of marketing to
the producers of dress materials leading to the development of
mass retailing of dress marketers. Also, this should be made in
a way requiring least external aid is of prime concern.
II. GENERAL DISCUSSION
A. Methodogy
The technique of Virtual Dressing Room for the virtual
fitting of clothes to a person involves the recognition of
human from the background with respect to light variations
and with least disturbance of other objects. This is to be
followed by detecting contour of both upper and lower body,
which is done by taking laplacian filter and then edge
detection. After then, feature points are extracted based on the
978-1-5090-3716-2/17/$31.00 ©2017 IEEE
Vivek S.
Department of Electronics and Instrumentation Engineering
Sri Ramakrishna Engineering College
Coimbatore, India
vivekgreen01@gmail.com
basic structure of human[5]. With these points as reference the
sample shirt is warped to fit for the person perfectly.
B. Expected benefits to the target groups
This algorithm mainly aims for men’s wear. Generally men’s
wear has to be selected very carefully as ones clothing speaks
very much about the person before they speak. And also the
choice of matching the shirt and pants is of peculiar concern.
The benefit of bringing the wide choice of online availability
with the assistance of feeling how cloths would look, to the
hands of person through the technology gives a great pleasure
to the customers.
This Virtual dressing room also aims at costume designers
to check how cloths looks before the actual manufacturing
enabling one to try new more designs of both formals and
casuals for men leading to a new trend set up and bringing out
of wide and different products to the market.
III. EXISTING METHOD
Existing virtual fitting approaches can be divided into two
main groups, 3D model based and 2D image based [12]. There
are certain systems that produce results of 3D simulations of
the human model [1,9] and cloth using the physical parameters
of the garments.In recent dressing simulations that can
reproduce detailed drapes or folds of garments fit on various
different body shapes. In the above systems simulations are
done with the help of a predefined human model
pictures.Customer’s 3D model can be generated using depth
camera. Virtual dressing room solutions work by overlaying
the 3d model or picture with in the customers live video
feed.In the video view the customer can feel garment or the
accessory virtually according to the movement of the customer
with the help of superimposed 3D model [6] or picture. Real
3D simulation fitting rooms has the features of both 3D
solutions and photo accurate fitting. With the photo and the
simple measurements of the body a 3D figurine is generated,
which accurately visualizes customer in chosen apparel items.
Another technique for foreground extraction is by complex
algorithms such as using grab-cut by obtaining adaptive
update tri-map. Other method for contour extraction is the
background subtraction by frame to frame subtraction which
requires fixed camera[16,17]. Virtual dressing room using 2
IEEE International Conference on Computer, Communication and Signal Processing (ICCCSP-2017)

dimensional images [3,13] requiring high contrast and good E. Virtual fitting
light settings have been developed. The next process is to superimpose the warped shirt over
IV. PROPOSED METHOD the person and to adjust the errors which may occur during the
imposing process due to mismatch in hand size, overlay[12] of
A. Foreground extraction old shirt, hip correction and kind of. Firstly the shirt is made to
The foremost step in the process of virtual fitting is to impose over the person after warping by the process of image
extract foreground from background. The algorithm works fusion[10] by blending as shown in Fig. 1(e). The error
well on varying light intensities provided that the background adjustments are done by applying the skin matrix over the
has got rid of presence of highly projecting bodies. For difference in sample shirt overlaid and the original images.
reliable extraction, before edge detection, colour clustering by Finally, the virtually dressed person is fused [7] to the
k means is performed. K-means clustering [2] basically background again, which can be the old background or a new
converts the image into clusters by quantifying the visual one for the aesthetic view of virtual dressing.
difference according to luminosity and chromaticity [14] as
V. SOFTWARE USED
shown in the Fig. 1(a). By doing so, the background is
perfectly grouped to a separate cluster. Pre-processing of The above algorithm has been realized with the help of
image is the delicate process which decides the reliability of MATLAB Software. As the software deals with handling of
further output. Hence, after these, pre-processing the image matrices, MATLAB is the perfect choice for dealing with
such as to remove small objects and morphological dilation is images as every pixel can be considered to be a matrix
done. element and hence, the programming is quite easy.
Additionally, Image processing toolbox provides with many
B. Extraction of human silhouette number of library files which can be directly utilized for
Extraction of human silhouette from the image in a reliable processing. Also, software is fast processing i.e. it could
and efficient way by riding off the disturbances by background process 1000*1000 pixels in one second.
is the crucial step for virtual fitting. Background separation
process is done by creating a morphological structuring
element for applying laplacian filter. Edge detection VI. RESULTS
Algorithm is employed to detect the edges of human. Then to
obtain the perfect single line silhouette of human contour,
traversal along two axes towards the centre has been carried
out. After all, post-processing involves dilation, closing the
contour are been performed as in Fig. 1(b).
C. Feature points Extraction
Feature points extraction is based on the much known
morphological facts of human structure. The shoulder and hip
points have to be found perfectly and precisely which plays a
vital role in accurate fitting of attire. This involves the fact that
at the shoulder of human body has the maximum slope (a)
considering below the face. So, before doing so, face is
detected by Viola-Jones Algorithm[15]. Another fact, being
the hip of the human has the maximum width is employed
here. Hip points are obtained by finding the maximum of the
single line silhouette arrayas shown in Fig. 1(c). This is
performed on both sides from the centre giving the reference
points for warping.
D. Warping of shirt
The sample shirt selected for fitting must be resized
according to the size of the person. Thus the sample shirt is (b)
warped to the silhouette by taking feature points as reference.
The Fig. 1(d) clearly shows the warping of shirt. This part of
work requires tedious algorithm for fine results. Warping is
the technique by which image is made to resize and reshaped
such that one images superimposes over another image of
similar contour[8]. The shirt is divided to regions of
distinguished variations and each being resized and
concatenated over which interpolation technique is performed.

978-1-5090-3716-2/17/$31.00 ©2017 IEEE

IEEE International Conference on Computer, Communication and Signal Processing (ICCCSP-2017)

REFERENCES
[1] K.Srinivasan, K.Porkumaran, G.Sai Narayanan, “Intelligent human body
tracking, modelling and activity analysis of video surveillence system:A
Survey”,Journal of convergence in engineering, technology and science,
Vol.1,pp.1-8,2009.
[2] Max Mignotte,“Segmentation by Fusion of Histogram based K-Means
Clusters in different color space”IEEE Transactions on Image
Processing, Vol.17,pp.780-787,2008.
[3] F. Cordier, W. Lee, H. Seo, and N. Magnenat-Thalmann, “From 2D
Photos of Yourself to Virtual Try-on Dress on the Web”, Springer, pp.
31–46,2001.
[4] D. Protopsaltou, C. Luible, M. Arevalo-Poizat, and N. Magnenat-
Thalmann, “A body and garment creation method for an internet based
(c) virtual fitting room”, in Proc. Computer Graphics International 2002
(CGI ’02). Springer, pp. 105–122,2002.
[5] F. Cordier, H. Seo, and N. Magnenat-Thalmann, “Made-tomeasure
technologies for an online clothing store”,IEEE Comput. Graph. Appl.,
vol. 23, no. 1, pp. 38–48, Jan. 2003.
[6] K.Srinivasan, K.Porkumaran, G.Sai Narayanan,”Skin colour
segmentation based 2D and 3D human pose modelling using Discrete
Wavelet Transform”,Journal of Pattern recognition and image
(d) Analysis,Springer,Vol.21,pp.740-753,2011.
[7] R. Brouet, A. Sheffer, L. Boissieux, and M.-P. Cani, “Design preserving
garment transfer”,ACM Trans. Graph., vol. 31,No. 4, pp. 36:1–36:11,
Jul. 2012.
[8] W. Xu, N. Umentani, Q. Chao, J. Mao, X. Jin, and X. Tong,“Sensitivity-
optimized rigging for example-based real-time clothing synthesis”,ACM
Trans. Graph. (Proc. of SIGGRAPH 2014), Vol. 33, No. 4, Aug. 2014.
[9] J. Tong, J. Zhou, L. Liu, Z. Pan, and H. Yan, “Scanning 3D full human
bodies using kinects”,IEEE Transactions on Visualization and Computer
Graphics (Proc. of IEEE Virtual Reality), Vol. 18, No. 4, pp. 643–650,
(e) 2012.
[10] J. Ehara and H. Saito, “Texture overlay for virtual clothing based on pca
Fig. 1. Different stage involved in the virtual dressing.(a)Colour of silhouettes”, in Proceedings of the 5th IEEE and ACM International
clustering. (b)Contour extraction. (c)Feature points extraction. (d) Shirt Symposium on Mixed and Augmented Reality, ser. ISMAR ’06. IEEE
warping. (e) Virtual fitting. Computer Society, pp.139–142,2006.
[11] Z. Zhou, B. Shu, S. Zhuo, X. Deng, P. Tan, and S. Lin, “Image-based
VII. CONCLUTION AND FUTURE WORK clothes animation for virtual fitting”, in SIGGRAPH Asia 2012
Technical Briefs, pp. 33:1–33:4,2012.
Growth in online shopping and the wish of people to have to
[12] P. Guan, O. Freifeld, and M. J. Black, “A 2D human body model
enjoy its maximum utilization on purchase of dress with dressed in eigen clothing”, in Proceedings of the 11th European
complete satisfaction of personal realization justifies the need Conference on Computer Vision: Part I, pp. 285–298,2010.
to develop an algorithm which virtually dresses people with [13] Y. Weng, W. Xu, Y. Wu, K. Zhou, and B. Guo, “2D shape deformation
the selected dress. In order to achieve the above requirement, using nonlinear least squares optimization”,Vis. Comput., vol. 22, no. 9,
the algorithm in such a way that everything works reliably pp. 653–660, 2006.
without the aid of external light adjuster or 3-D viewing or [14] S. Bianco and R. Schettini, “Color constancy using faces”, in 2012 IEEE
Conference on Computer Vision and Pattern Recognition (CVPR), pp.
fixed camera, which is usually needed for the present
65–72,2012.
algorithms creating the limitations of its own use, has been
[15] Guan-Chun Luh “Face detection using combination of skin colour pixel
developed to reach out to all people. This algorithm would detection and viola jones face detector”, in 2014 IEEE International
enable people to check out themselves with different dresses Conference on Machine Learning and Cybernetics(ICMLC),pp.364-370,
with less number of restrictions will give great pleasure to the 2014.
online shoppers. [16] K.Srinivasan, K.Porkumaran, G.Sai Narayanan, “Background
In future, the algorithm is planned to be modified to find the subtraction techniques for human body segmentation in indoor video
survelliance”,Journal of Scientific and Industrial
human silhouette with variable background and noisier Research,Vol.73,pp.342-345,2014.
environment. Which is the more challenging task in still image [17] K.Srinivasan, K.Porkumaran, G.Sai Narayanan, “Enhanced Background
using imageprocessing. Subtraction Techniques for Monocular Video
Applications”,International Journal of Image Processing and
ACKNOWLEDGMENT Applications,Vol. 1,no. 2,pp.87-93,2010.
The authors would like to sincerely thank the Management,
Director and Principal of Sri Ramakrishna Engineering
college, Coimbatore for their constant support and providing
us with the required facility.

978-1-5090-3716-2/17/$31.00 ©2017 IEEE