Human Vision Inspired Based Image Illumination
Enhancement by Using Local Singular Value
Decomposition and Discrete Wavelet Transform
Gholamreza Anbarjafari
iCV Group, IMS Lab, Institute of Technology, University of Tartu,Tartu, Estonia, email: sjafari@ut.ee
Cagri Ozcinar
Department of Electronic Engineering, University of Surrey,GU2 7XH,
Surrey, UK email: cagri.ozcinar@ieee.org
Hasan Demirel
Department of Electrical and Electronic Engineering, Eastern Mediterranean
University, Gazimağusa, KKTC, Mersin 10, TURKEY, email: hasan.demirel@emu.edu.tr
Abstract: Recently, many computer vision applications are being inspired by human behavior, or human visual system. Also it is known that illumination issues have always been an important problem in many image processing applications. In this work we propose a new image illumination enhancement technique which is inspired from the human visual system behavior on illumination correction. The proposed technique uses local singular value decomposition (SVD) and discrete wavelet transforms (DWT), inspired from the fact that human visual system equalizes a scene by disregarding the extreme illuminated areas. In other words, human brain uses local illumination enhancement and this localization is based on the extreme illuminations, e.g. existence or absence of too much light. In this technique, after dividing the image into several locals, each local is converted into the DWT domain and after updating the singular value matrix of the respective low-low subband, the local is reconstructed by using inverse DWT (IDWT). Combination of locals results in the equalized image. The technique is compared with the standard general histogram equalization (GHE) and local histogram equalization (LHE). The experimental results are showing the superiority of the proposed method over the aforementioned techniques.
Keywords: Illumination enhancement, Discrete wavelet transform, Singular value decomposition, Image processing
I.
I
NTRODUCTIONNowadays, computer graphics applications are being developed rapidly by inspiring from human vision system (HVS). Usually human vision inspired (HVI) techniques perform better than the previous techniques, since our brain performs many computer vision applications
magnificently. Therefore developing an image
illumination enhancement inspired from HVS can result in a successful technique.
Contrast enhancement is frequently referred as one of the most important issues in image processing. Contrast is created by the difference in luminance reflected from two adjacent surfaces. In other words, contrast is the difference in visual properties that makes an object distinguishable from other objects and the background. In visual perception, contrast is determined by the difference in the color and brightness of the object with other objects. Our visual system is more sensitive to contrast than absolute luminance; therefore, we can perceive the world similarly
regardless of the considerable changes in illumination conditions.
If the contrast of an image is highly concentrated on a specific range, e.g. an image is very dark; the information may be lost in those areas which are excessively and uniformly concentrated. The problem is to optimize the contrast of an image in order to represent all the information in the input image. There have been several techniques to overcome this issue [1-4].
When we are looking at an object, whose background is filled with huge amount of light, e.g. sun exist in the background, our eye observe the object in a very low illumination and contrast. What we, as a human, usually do in such scenario is as follows: we use our hand in order to block the high illumination sections (e.g. sun), then the remaining scene will be automatically equalized by our visual system. Due to absence of the extreme illumination the scenes will be observed with a good illumination. In our daily life we use this action very frequently. In this work we are proposing to use this task in order to divide an image into several locals. One of the main contributions of this work is actually the division of the image into the HVI based locals. Then these locals will be equalized by using our previously introduced technique which is based on singular value decomposition (SVD) and discrete wavelet decomposition (DWT) [5].
In the equalization stage there exist many techniques. One of the most frequently used techniques is general histogram equalization (GHE). After the introduction of GHE, researchers came out with better techniques such as local histogram equalization (LHE). In LHE, the locals are simply obtained by dividing the image into several sub-images, e.g. 16x16. Nowadays there are many new techniques for illumination enhancement such as dynamic histogram equalization (DHE) [6] and singular value equalization (SVE) [7].
value matrix obtained by singular value decomposition (SVD). SVD of an image, which can be interpreted as a
matrix is written as follows:
Fig.1:A grey scale image and generated image by manipulating several singular values.
T
A
A A
A
U
V
(1)where UA and VA are orthogonal square matrices known
as hanger and aligner respectively, and ΣA matrix contains
the sorted singular values on its main diagonal. The idea of using SVD for image equalization comes from this fact
that ΣA contains the intensity information of the given
image [13]. Now the question is which of the singular values has the highest effect on illumination. For this
purpose σ1 and other singular values have been changed in
order to see the effect of change of singular values on illumination of the images. Fig. 1 is representing a grey scale image and the generated images by manipulating singular values.
It can easily be observed from Fig. 1 that a significant change on the illumination is possible if the first singular
value, σ1, is changing. Hence manipulating the biggest
singular value implies the idea of using it in illumination correction. In [7], the method uses the ratio of the largest singular value of the generated normalized matrix over a normalized image which can be calculated according to equation (2).
0,var 1
max
max
N
A
(2)
whereN0,var 1is the singular value matrix of the
synthetic intensity matrix. This coefficient can be used to regenerate an equalized image using equation (3).
T
equalized
A
U
A
A
V
A
(3)whereequalizedAis representing the equalized image A.
This task is eliminating the illumination problem.
Nowadays, wavelets have been used in image processing frequently. It has been used for feature extraction [13], denoising [14], compression [15], face recognition [16], and illumination equalization [5]. The decomposition of images into different frequency ranges permits the isolation of the frequency components
introduced by “intrinsic deformations” or “extrinsic factors” into certain subbands [17]. This process results in
isolating small changes in an image mainly in high frequency subband images. Hence discrete wavelet transform (DWT) is a suitable tool to be used for designing pose invariant face recognition system. The DWT results in four decomposed subband images refer to Low-Low (LL), Low-High (LH), High-Low (HL), and High-High (HH). The frequency components of those subband images cover the frequency components of the original image.
In this work, we have proposed a new method for image illumination enhancement which is inspired from HVS in order to divide the image into several locals and then each local is equalized by using our previously introduced technique [5] which is based on SVD of LL subband image obtained by DWT. In this paper, the proposed method has been compared with GHE technique as well as LHE. The results indicate the superiority of the proposed HVI based local SVE+DWT illumination enhancement technique performs better than the mentioned methods.
1st is 0 2nd is 0 3rd is 0
4th is 0 5th is 0 3rd and 2nd s are 0
II.
T
HEP
ROPOSEDI
MAGEI
LLUMINATIONE
NHANCEMENTIn this technique the image is divided into two equal sections. Then the mean of each local is calculated. The local with higher mean will be divided into another two sections. This step will be repeated until the mean of one of the locals is greater than a given high boundary threshold value, e.g. 200 for a gray scale 8-bit image representation. Then this local is referred as high intensity local. Similar steps will be done for the locals with the lower mean till the local with mean lower than a given lower boundary threshold is achieved. This local is referred as low intensity local. This division of the image into three locals (high intensity local, low intensity local, and the remaining part) is actually inspired from the HVS, as it was mentioned in the introduction. Now each local is equalized by using SVE+DWT technique [5] which can be explained as follows: As it was mentioned, the singular value matrix obtained by SVD contains the illumination information. The biggest singular value has the highest impact on the illumination of the image. Therefore, changing singular values will directly affect the illumination of the image hence the other information in the image will not be changed. Also it was mentioned that the illumination information is embedded in LL subband. The edges are concentrated in other subbands (i.e. LH, HL, and HH). Hence, separating the high frequency subbands and applying the illumination enhancement in LL subband only, will protect the edge information from possible degradation. After reconstructing the final image by using IDWT, the resultant image will not only be enhanced with respect to illumination, but also it will be sharper.
The general procedure of the proposed technique is as
follows. The input local, A, is first processed by using
GHE to generate Â. Then both of these images are
transformed by DWT into four subbands. The correction coefficient for singular value matrix is calculated by using the following equation:
max maxˆ
LL
LL
A
A
(4)whereLL
A
is the LL singular value matrix of the inputimage and LLˆ
Ais the LL singular value matrix of the
output of the GHE. The new LL local is composed by:
LL LL LL LL LLA
A
A
LL
U
A
A
V
A
(5)
Now the LLAand LHA, HLA, and HHAsubband images
of the original local are recombined by applying IDWT, to
generate the resultant equalized local
A
.
A,
A,
A,
AA
IDWT LL LH HL HH
(6)In this work we have used db.9/7 wavelet function as the mother function of the DWT. In the following section the experimental results and the comparison of the aforementioned techniques are discussed. Fig. 2 illustrates all steps of the proposed image equalization technique.
low intensity local Equalized local using GHE DWT DWT LL LH HL
HH LL LH HL HH
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate ζ
using eqn (4) Calculate the new Σ and reconstruct the new LL image, by using eqn (6). IDWT Equalized local remaining locals Equalized local using GHE DWT DWT LL LH HL
HH LL LH HL HH
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate ζ
using eqn (4) Calculate the new Σ and reconstruct the new LL image, by using eqn (6). IDWT Equalized local high intensity local Equalized local using GHE DWT DWT LL LH HL
HH LL LH HL HH
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate the U, Σ, and V for LL subband image and find the maximum element in Σ.
Calculate ζ
using eqn (4) Calculate the new Σ and reconstruct the new LL image, by using eqn (6). IDWT Equalized local Input image
Equalized image
III.
E
XPERIMENTALR
ESULTS ANDD
ISCUSSIONSFig. 3, 4, and 5 (a) illustrate low contrast face images from Yale face database [18]. These images have been
equalized by using GHE (b), HSI based localized GHE (c), and HSI based localized SVE+DWT (d). The quality of the visual results indicates that the proposed equalization technique is sharper and brighter than the one achieved by other techniques.
Fig. 3: (a) Low contrast image, Equalized image using: (b) GHE, (c) HSI based localized GHE, and (d) the proposed technique.
Fig. 4: (a) Low contrast image, Equalized image using: (b) GHE, (c) HSI based localized GHE, and (d) the proposed technique.
(a)
(d) (b)
(c)
(a)
(d) (b)
Fig. 5: (a) Low contrast image, Equalized image using: (b) GHE, (c) HSI based localized GHE, and (d) the proposed technique.
IV.
C
ONCLUSIONIn this work, a new HVI based image illumination enhancement technique using SVD and DWT was proposed. The proposed technique divided the image into three locals, and then converted each local from spatial domain into the DWT domain and after equalizing the singular value matrix of the LL subband; it reconstructed the local in the spatial domain by using IDWT. Afterwards the locals are being combined in order to generate the image. The experimental results were showing the superiority of the proposed method over the conventional techniques.
A
CKNOWLEDGMENTAuthors would like to thank Prof. Dr. Ivan Selesnick from Polytechnic University for providing the DWT codes in MATLAB.
This work is partially supported by ERDF program
“Estonian higher education information and
communications technology and research and
development activities state program 2011-2015 (ICT
program)”
R
EFERENCES[1] S. Chitwong, T. Boonmee, and F. Cheevasuvit, ‘Enhancement of colour image obtained from PCA-FCM technique using local area
histogram equalization’, Proc. SPIE, 2002, 4787, pp. 98–106. [2] W. G. Shadeed, D. I. Abu-Al-Nadi, and M. J. Mismar, “Road traffic
sign detection in color images”, 10thInternational Conference on
Environmental and Computer Science, Vol. 2, Dec 2003, pp.
890-893.
[3] R. C. Gonzalez, and R. E. Woods, Digital Image Processing, Prentice Hall, ISBN 013168728X, 2007.
[4] T. K. Kim, J. K. Paik, and B. S. Kang, “Contrast enhancement system using spatially adaptive histogram equalization with
temporal filtering,” IEEE Transactions on Consumer Electronics, Vol. 44, No. 1, 1998, pp. 82–86.
[5] C. Ozcinar, H. Demirel, and G. Anbarjafari, "Satellite Image Contrast Enhancement Using Discrete Wavelet Transform and Singular Value Decomposition", IEEE Geoscience and Remote
Sensing Letter, April 2010, Vol. 7, No. 2, pp. 334-338.
[6] M. Abdullah-Al-Wadud, H. Kabir, M. A. A.Dewan, C. Oksam, “A
dynamic histogram equalization for image contrast enhancement”,
IEEE Transaction on Consumer Electronics, Vol. 53, No 2, May
2007, pp. 593-600.
[7] H. Demirel, G. Anbarjafari, and M. N. S. Jahromi, "Image equalization based on singular value decomposition", 23rd IEEE
International Symposium on Computer and Information
Sciences, Turkey, Oct 2008, pp. 1-5.
[8] T. Kim, H.S. Yang, “A Multidimensional Histogram Equalization by Fitting an Isotropic Gaussian Mixture to a Uniform
Distribution”, IEEE International Conference on Image Processing, 8-11 Oct. 2006, pp. 2865 – 2868.
[9] A. R. Weeks, L. J. Sartor, and H. R. Myler, ‘Histogram specification of 24-bit color images in the color difference (C-Y)
color space’, Proc. SPIE, 1999, 3646, pp. 319–329.
[10] F. Kahraman, B. Kurt, M. Gokmen, “Robust Face Alignment for
Illumination and Pose Invariant Face Recognition”, IEEE Conference on Computer Vision and Pattern Recognition, 2007. CVPR '07, pp. 1 – 7.
[11] C. C. Sun, S. J. Ruan, M. C. Shie, and T, W. Pai, “Dynamic
contrast enhancement based on histogram specification”, IEEE
Transactions on Consumer Electronics, Vol. 51, No. 4, 2005,
pp.1300–1305.
[12] G. Anbarjafari, “Face recognition using color local binary pattern
from mutually independent color channels”,EURASIP Journal on
Image and Video Processing, 2013.
[13] J. W. Wang and W. Y. Chen, “Eye detection based on head contour geometry and wavelet subband projection”, Optical Engineering, 2006, Vol. 45, No. 5.
[14] J. L. Starck, E. J. Candes, and D. L. Donoho, “The curvelet
transform for image denoising”, IEEE Transactions on Image
Processing, 2002, Vol. 11, pp: 670-684.
(a)
(d) (b)
[15] A. M.Rufai, G.Anbarjafari, and H.Demirel, “Lossy image compression using singular value decomposition and wavelet
difference reduction”,Digital Signal Processing, 2014, Vol. 24, pp. 117-123.
[16] C. C. Liu, D. Q. Dai, and H. Yan, “Local discriminant wavelet
packet coordinates for face recognition”, Journal of Machine
Learning Research, 2007, Vol. 8, pp: 1165-1195.
[17] D. Q. Dai and H. Yan, “Wavelet and face recognition”, Face
recognition, Chapter 4, K. Delac and M. Grgic (Eds), ISBN:
978-3-902613-03-5, Austria, pp: 59-74, 2007.
[18] A. S. Georghiades, P. N. Belhumeur, and D. J. Kriegman, “From few to many: illumination cone models for face recognition under
variable lighting and pose”, IEEE Transaction on Pattern Analysis
and Machine Intelligence, 2001, Vol. 23, pp. 643-660.
A
UTHOR’
SP
ROFILEGholamreza Anbarjafari received his B.Sc., M.Sc., and Ph.D. degrees from Department of Electrical and Electronic Engineering at Eastern Mediterranean University (EMU), Famagusta, North Cyprus, in 2007, 2008, 2010 respectively. He has been working in the field of image processing and is currently focusing in many research works related to multimodal emotion recognition, image illumination enhancement, super resolution, image compression, watermarking, visualization and 3D modelling, and computer vision for robotics. He has been involved in several human-robot interaction projects. He is head of iCV group and is currently working as an Assoc. Prof. in IMS Lab, Institute of Technology at University of Tartu.
Cagri Ozcinar received the B.Sc. degree in electrical and electronics engineering in 2009, and the M.Sc. degree in electronic engineering in 2010. He is currently pursuing the Ph.D. degree in the electronic engineering at the I-Lab Multimedia Communications Research Group, Centre for Vision, Speech, and Signal Processing (CVSSP), University of Surrey, Guildford, UK. His research interests include 3D/multi-view video compression, processing and delivery techniques.
Prof. Dr. Hasan Demirel received his PhD and MSc degrees from Imperial College Science, Technology and Medicine in 1998 and King's College London in 1993 respectively after graduating from the Electrical and Electronic
Engineering Department of Eastern