Proceedings of the 4th International Symposium on Communications,

Control and Signal Processing, ISCCSP 2010, Limassol, Cyprus, 3-5 March 2010

Modified Histogram Equalization for Image Contrast Enhancement

Kota Murahira, Takashi Kawakami, and Akira Taguchi*), Member IEEE

Abstract- Histogram equalization (HE) is one of the common for HE is derived by the cumulative histogram. In our method,
methods used for improving contrast in digital images. However, first, the histogram is modified. Next, the transformation
this technique will cause an effect on brightness saturation or function is derived by the modified histogram. We can also
shadow in some almost homogeneous area. In order to avoid express the transformation function of the HEwVED by using
these effects, mean preserving bi-histogram equalization (BBHE)
the transformation function of the HE. In the proposed HE
method is proposed. In this paper, we propose a novel HE with
variable enhancement degree (HEwVED). Moreover the BBHE with variable enhancement degree, the emphasis level can be
with variable enhancement degree (BBHEwVED) is also changeable by operating one parameter.
proposed. The enhancement degree of each of these methods is This paper also presents the modified BBHE. In the BBHE,
controlled by only one parameter. We also show the guide to the histogram is separated by the mean brightness value. Thus,
decide this parameter. Since our method is able to control the images with considerable high or low mean brightness value
enhancement degree and the mean of image is tend to be cannot be emphasized efficiently, since the range of the one
preserved, all kinds of images are enhanced adequately. separated part becomes very narrow. Then, the BBHE with
variable enhancement degree (BBHEwVED) is proposed. The
I. INTRODUCTION enhancement degree ofBBHEwVED is also controlled by one
Image enhancement is a process involving changing the parameter.
pixels' intensity of the input image, so that the output image The enhancement degree of each of two proposed methods
should subjectively looks better [1]. The purpose of image is controlled by one parameter. We offer the guidance in how
enhancement is to improve the interpretability or perception to decide parameter. Experimental results show the
of information contained in image for human viewers. effectiveness of the proposed method.
There are many image enhancement methods have been
proposed. A very popular technique for image enhancement is II. HISTOGRAM EQUALIZATION WITH VARIABLE
histogram equalization (HE). This technique is commonly ENHANCEMENT DEGREE
employed for image enhancement because of its simplicity Conventional histogram equalization (HE) for digital
and comparatively better performance on almost all types of original images is explained as follows. Next we propose the
images. The operation of HE is performed by remapping the novel HE with variable enhancement degree.
gray levels of the image based on the probability distribution
of the input gray levels. It flattens and stretches the dynamic A. Histogram Equalization
range of the image's histogram and resulting in overall Suppose original image g(i, j) composed of discrete gray
contrast enhancement. levels in dynamic range of [0, L-l]. The histogram of the
However, some images are produced undesirable artefacts original image g(i, j) is denoted h(x) (0 ~ x ~ L -1). x is the
by HE. One of the reasons this problem is because HE
normally changes the brightness of the input image pixel gray level. Let the number of pixels of g(i, j) is N.
significantly, makes some of the uniform regions of output According to the histogram equalization (HE), the gray level x
image become saturated with very bright or very dark is converted to y as follows.
intensities. Thus, it is necessary for the HE to control the
enhancement degree. One solution is the dynamic HE [2]-[5]. Y = fHE(X) = rH(x).(L-l)/ Nl (1)
The dynamic HE can maintain the original input brightness in
where H(x) is cumulative histogram of the range from 0 to x.
the output image. The earliest work which is belonged to the
dynamic HE, has been proposed by Kim [2]. This method is
r.l denotes the rounding operation. fHE(X) means the
called the brightness preserving bi-histogram equalization transformation function of the HE.
(BBHE). In this method, the image histogram is divided into
B. Histogram Equalization with Variable Enhancement
two parts. The separation intensity is presented by the input
mean brightness value. After this separation process, these
Degree
two histograms are independently equalized. The HE degrades some images, which have large
In this paper, we propose a novel HE with variable background with high or low intensity. First, we propose HE
enhancement degree (HEwVED). Transformation functions with variable enhancement degree (HEwVED).
The histogram of input image h(x) is modified as
The authors are with the Tokyo City University, 1-28-1 Tamazutsumi,
Setagaya-ku, 158-8557, Tokyo Japan.
*) E-mail: [email protected] ha(x) =a{h(X)- ~}+ ~ (2)

978-1-4244-6287-2/10/$26.00 ©2010 IEEE

where NIL shows the average frequency for one gray level.
Gray level x is transformed to Y a as

(3)

Eq.(3) can rewrite as the following expression.

Ya =fHEwVED(X) =a -fHE(x)+(l-a) ·x (4)
(a) a =0 (original image)
fHEwVED(X) means the transformation function of the HEwVED. ~
If a is set to 0, the enhanced image is equal to the original
image. On the other hand, if a is set to I, the HEwVED is
equivalent to HE. Thus, the HEwVED is changeable from
identity operation to HE.

C. Results
We use more than 30 images to evaluate the performance
of the HEwVED. We show the results of "couple" and "plan"
from these images. The histogram of images "couple" and
"plain" are biased low or high brightness shown as Fig.l .
We will use the following quantitative measures: mean
brightness (MB), the measure of enhancement (EME) and the
maximum gradient value of the transformation function
(MGTF). The measure of enhancement (EME) approximates
an average contrast in the image by dividing image into non- (e) a =0.8 (0 a = 1.0 (HE)
overlapping blocks, finding a measure based on minimum and
maximum intensity values in each block, and averaging them. Fig.2 Results ofHEwVED using "couple"
Fig.2 (a) and Fig.3 (a) show the original image "couple"
and "plan", respectively. These images have a wide
background with low or high intensity. Fig.2 (f) and Fig.3 (f)
show the results of applying the HE. A part of the background
is saturated or shadowed. Fig.2 (b)-(e) and Fig.3 (b)-(e) are
shown the results ofHEwVED.
In the case of "couple", the result ofHEwVED with a =0.6
looks very natural. On the other hand, HEwVED with a =0.4
shows best result for "plan". The MGTF of these two images
(b) a =0.2
is 4. We can understand the degree of the emphasis by MGTF.
Thus, a is coordinated so that MGTF becomes around 4.
The transformation functions are shown in Fig.4. In
"couple", the gradient of transformation function is very high
in low intensity region. In "plan", the gradient is very high in
high intensity region. These mapping makes unnatural results. '.
We can also understand that the shape of transformation
function ofHEwVED is preserved of that of HE.

:: ~
2000

1000

0 I
1 51 10 1 151 20 1 251 ~ 101 151 ~ 8 1

(a) couple (b) plan

Fig.l Histograms of "couple" and "plan"

Fig.3 Results ofHEwVED using "plan"

 parts . The separation intensity is the mean brightness of
Table I Quantitative measurement results ofHEwVED original image which is derived by Eq.(5)
The HE is applied to separately to the first sub-histogram in
MB EME MGTF the range of [0, X A -1] and second sub-histogram in the range
a couple plan couple plan couple plan of [XA • L-I] . Gray level x in the first range [0, X A -I] is
0.0 44 178 35 67 I I transformed to y as
0.2 59 167 48 78 2 3
0.4 75 156 61 89 3 4 Y=IH(x) ,(XA -1) / H 1 1 (6)
0.6 92 146 74 100 4 5
0.8 108 135 88 112 7 7 where
1.0 124 125 101 123 9 8
Xrl
HI = Ih(n) (7)
300 r--- - - - - - - - - - - - - ---, n=O

Gray level x in the second range [XA , L-l] is transformed to

...r
200 f--- -
x2 as
100 f -- IF-- ---=. -"""'--- - ---1 (8)

o .-.r:::.;-;..wawwWWWID.wwwowcwwww""",,""wwwowwluwWiWiiiiiWii i i WWiiiiiiwI where

1 51 101 151 201 251
L-I

(a) couple H 2 =N -HI = Ih(n) (9)

n=X A

300
We integrate Eq.(6) with Eq.(8) and make the
200 j--L-~~~------:::;;;j~F-------l transformation function fBBHE(X), Thus, In the case ofBBHE,
gray level x is transformed to Y as
100 1-----~"''=_..-=:------c::iO~-______1

y=fBBHE(X) (10)

51 101 151 201 251

B. Brightness Preserving Bi-histogram Equalization with
Variable Enhancement Degree
(b) plan
It is important to preserve the mean brightness of image for
the consumer electronics. However, for photograph, the
FigA The transformation functions ofHEwVED
change in a few mean values is permitted. Thus , we also
propose the BBHE with variable enhancement degree
III. BRIGHTNESS PRESERVING BI-HISTOGRAM
(BBHEwVED).
EQUALIZATION WITH VARIABLE ENHANCEMENT DEGREE
The transformation function of BBHEwVED is defined by
The HE shows the high performance in enhancing contrasts the following:
of a given image, however, it is rarely employed in consumer
electronics such as TV since the straight use of histogram fBBHEwVED(X) ={3 -fBBHE(X) +(1- {3) . fHE(X) (11)
equalization may change the original brightness of an image,
If {3=0, the BBHEwVED is equivalent to the HE. On the
deteriorate visual quality, or, introduce some annoying
other hand, if {3=I, The HEwVED is equivalent to BBHE.
artifacts. If the mean brightness of the enhanced image is
same as original image, the abovementioned problem is Thus, the BBHEwVED is changeable from HE to BBHE.
resolved. One of mean brightness preserving method is the
brightness preserving bi-histogram equalization (BBHE). C. Results
In this section, BBHE is explained briefly and BBHE with In order to evaluate the performance of BBHEwVED, we
variable enhancement degree (BBHEwVED) is proposed. also use "couple" and "plain". The results of BBHEwVED
A. Brightness Pres erving Bi-histogram Equalization using "couple" and "plan" are shown in Fig.5 and Fig.6,
respectively.
The mean brightness X A of original image is calculated by
Fig.5(t) and Fig.6(t) show the results of applying the
BBHE. In the case of "plan", the mean brightness is preserved
(5)
by processing BBHE. In the case of "couple", the mean
brightness of the result of BBHE is slightly different from the
In the brightness preserving bi-histogram equalization original image. However, we don't feel the change visually
(BBHE), the original image histogram is divided into two
 Table 2 Quantitative measurement results ofHEwVED

MB EME MGTF
P couple plan couple plan couple plan
0.0 124 125 101 123 9 8
0.2 113 135 99 119 8 7
0.4 103 146 97 115 7 6
0.6 92 157 96 112 6 5
0.8 81 168 94 108 5 5
1.0 71 179 92 104 5 4
original 44 178 35 67 I I

from a processing result. Both "couple" and "plan", the

contrast enhancement degree is insufficient.
Fig.5 (b)-(e) and Fig.6 (b)-(e) are shown the results of
BBHEwVED. In the case of "couple", we can fell the change
visually when the mean brightness setting P is less than 0.4.
P=0.6 or 0.8 is suitable for "couple". In the case of "plan",
P=O.4 or 0.6 is suitable. In these cases, the MGTF of images
is 5 or 6. Thus, fJ is coordinated so that MGTF becomes
around 5.
Finally, transformation functions are shown in Fig.7. The
parameter P is set to 0.6. The characteristic of the curve of
BBHEwVED is inherited of that of BBHE. We can
understand that the BBHEwVED generates various
transformation functions by changing P
Fig.5 Results ofBBHEwVED using "couple"

In this paper, we have proposed the HEwVED and the

BBHEwVED. The HEwVED is an extended version of HE.
The HEwVED can be controlled the enhancement degree.
The BBHEwVED is an extended version of BBHE. In order
to emphasize all kind of images efficiently, we permit some
different the mean brightness between the original image and
the enhanced image. We also illustrate the suitable value for
two parameters (Le., a and {3).
From experimental results, we can understand that
especially, BBHEwVED can realize the natural enhancement.

REFERENCES

[1] Rafael C. Gonzalez, and Richard E. Woods, "Digital Image

Processing", 2nd edition, Prentice Hall, 2002.
[2] Yeong-Taeg Kim, "Contrast enhancement using brightness
preserving bi-histogram equalization", IEEE Trans. Consumer
Electronics, vol. 43, no.l, pp.l-S, Feb. 1997.
[3] Soong-Der Chen, and Abd. Rahman Ramli, "Contrast
enhancement using recursive nean-separation histogram
equalization for scalable brightness preservation", IEEE Trans.
Consumer Electronics, vol. 49, no.4, pp.1301-1309, Feb. 1997.
[4] M. A-A-Wadus, M. H. Kabir, M. A. Dewan, and o. Chae, "A
dynamic histogram equalization for image contrast
enhancement", IEEE Trans. Consumer Electronics, vol. 53, no.2,
pp.593-600, May 2007
[5] H. Ibrahim, and N. S. P. Kong, "Brightness preserving dynamic
histogram equalization for image contrast enhancement", IEEE
Trans. Consumer Electronics, vol. 53, no.4, pp.1752-1758, Nov.
2007.