An improved hdr image processing using fast global tone mapping

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 02 Issue: 12 | Dec-2013, Available @ http://www.ijret.org 298
AN IMPROVED HDR IMAGE PROCESSING USING FAST GLOBAL
TONE MAPPING
Sajeena A1
, Sunitha Beevi K2
1
Assistant Professor, Department of Electronics and Communication Engg. TKM College of Engg. Kollam, Kerala,
India , sajina.nizam@gmail.com
2
Assistant Professor, Department of Electrical and Electronics Engg., . TKM College of Engg. Kollam, Kerala, India,
.sunitha_haris@yahoo.com
Abstract
People always try to reconstruct images that look like original scene. But it is not succeeded up to the expected level because
the display devices cannot accommodate the actual range of illumination. The original scene is having high dynamic range
(HDR) and the display devices are having low dynamic range (LDR) of values. Tone mapping can be used to display HDR image
in an LDR device. In this paper we are trying to develop a better tone mapping method. A number of images with different
exposure are fused to form an HDR image.. Here we used average method for this which is the simplest one. Then by using tone
mapping the HDR image is converted into LDR. We tried Tone Reproduction Curve (TRC) based global tone mapping. TRC
method results faster operation The global tone mapping operator algorithm is simple and does not introduce ghosting. The
image will also free from blur and halo like artifacts. We took log average of the images to keep the pixel values within limit.
Index Terms: Image fusion, High dynamic range images, , Tone mapping. Low dynamic range image, global
operator.
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
Digital cameras may give clear picture of the scene they are
capturing. With the increasing megapixel values of the
image capturing device (camera) the output become more
clearer. But they nevern succeeded in displaying the exact
dynamic range of illumination of the actual scene. This may
not be the fault of the camera alone but the common
displaying devises like CRT, LCD, LED displays, printers
etc cannot accommodate the wide dynamic range of the
scene. The real world scenes will have a dynamic range of
about ten orders of magnitude. This will vary according to
the climatic conditions and exposure time. Human eye can
preserve around five to nine orders of magnitude. The
display devices are having only two orders of dynamic
range. Even though most of the devices are following the
technology of human eye, they have not met the
requirement.
We can say that real world scenes are having High Dynamic
Range (HDR) of luminance values and the reproducing
devices are having Low Dynamic Range (LDR). High
Dynamic Range imaging allows capturing of the object with
a wide dynamic range. This may be done with the help of a
radiance map. Radiance map can be prepared in the
following way: Take a number of images with different
shutter speeds. Images obtained will have different
brightness levels. Darker portions will be more dark for
shorter exposure and longer exposures will make the
brighter areas more bright. Both the cases will make image
with less details
Radiance map is produced by fusing the multiple
exposure images [2] using suitable method. Here we used
the average method with a control factor ƒ. This factor will
control the output image brightness level. Methods for better
colour reproduction is also provided.
The resulted HDR image is subjected to tone mapping. Tone
mapping is the process of converting HDR images for
making it suitable for display on an LDR display device.
Two broad categories of tone mapping technology are
available They are tone reproduction curve (TRC) based and
tone reproduction operator (TRO) based, Pixel distributions
are considered in TRC based technique and TRO based
technique considered spatial manipulation of local
neighbouring pixel values,.
Spatial tone mapping can be done in two ways, either
globally or locally. Global operators compresses the image
with an identical curve on all pixels and local operators will
do it with different curves. They are done with different
scales.
In this project we tried [1] global tone mapping. We
took logarithmic intensity values of pixels to make it
between 0 and 1.
2. METHODOLOGY
Multiple exposure images can be combined to form an
HDR image with large dynamic range pixel values [2]. With
the increase in the number of images the dynamic range of

_______________________________________________________________________________________
the HDR can also be increased. Then a tone mapping
operator is applied to make the image displayed on an LDR
device.
A number of images of the scene (RAW images) is
taken with a digital camera by adjusting the shutter speed.
Cameras are available that can take multiexposure images
with a single click. Images From each image the red green
and blue pixel values are taken separately averaged are
combined with a weighting factor. The average value
determines the overall brightness of the image. Using this
algorithm it is possible to process about 21 images. Then a
global tone mapping algorithm is applied to compress the 32
bits of the HDR images to 8 bit LDR image.
The algorithm is implemented with Java NetBeans.
The block diagram is shown in figure 1
Figure 1 : Block diagram of HDR image processing
system
2.1 Combining multiple images:
Images can be ditectly taken using a digital camera or can
be downloaded fro the internet as jpeg images. They should
be taken with different shutter speeds. Let N be the number
of images of the scene taken.
A control parameter ƒ is taken to control the image
brightness. Here we took averaging method to make the
HDR image. It will help in building the HDR image with
average brightness of the overall images,
Including the ƒ coordinate value ,we need 32 bits to
represent an HDR image. That is 8 bits to represent blue, 8
bits to represent green, 8 bits to represent red and another 8
bits to represent ƒ value. We let the first 8 bits to represent
blue pixel value ,9 to 16 bits to represent green, 17 to 24 bits
to represent red and 25 to 32 bits to represent ƒ.
F F F F F F F F
1111 1111 1111 1111 1111 1111 1111 1111
ƒ Red Green Blue
Let βbe the bitrate.
Then the average value is calculated using the following
equation
Ravr =
Gavr =
Bavr =
Using these average values we constructed the HDR
image.In order to display this image in an LDR device we
used tone mapping.
2.2 Tonemapping:
Tone mapping is done to convert the 32 bit HDR image to
8 bit LDR image suitable for the display devices. It should
be done without loss of information.
1111 1111 1111 1111 1111 1111 1111 1111
1111 1111
We used an index whose value is from 0 to less than 3.
Let x be the length and y be the width of the matrix.
Therefore the number of pixels in the matrix Nn is x * y
Consider value „a‟ equal to a key value multiplied by the
brightness. It can be specified by the user.
Another constant µ is taken equal to 0.00001 for the log of a
black pixel in order to reduce the contrast.
For these values the log average luminance is calculated as

_______________________________________________________________________________________
Lavl =
Where Lw is the luminance of white. Now the scaled
luminance is calculated as
Scl =
Now the maximum luminance is mapped to that of pure
white.
Finally the tone mapping operator is applied.
The scaled luminance pixel value is calculated as
Scl [(1+ ) (1+Scl)]
These pixel values in the matrix form makes the LDR
image. The luminance value is always kept between zero
and one.
Here we are compressing the bit rate of the HDR image to
make it fit to an LDR display. That is the information
contained in 32 bits should be represented by 8 bits. This
will reduce the clarity of the picture.
In this project we can fuse multiexposure images from three
to twenty one .More number of images will make the output
more clearer.
3. RESULTS
The following multexposue images are processed.
These images after processing gave the following output

_______________________________________________________________________________________
Another example. Here we took four different exposures
Images 1 to 4
The LDR image obtained is
In this paper we used Java Netbeans for doing the operation
Tonemapping operators are applied on the images using the
above mentioned formula].We took {1},[2}, [3],[4], [5], [6]
as the references. We also took [7]. [8], [9] for the HDR
image and colour correction.
C. Abbreviations and Acronyms
HDR High Dynamic Range Images
LDR Low Dynamic Range Images
TRO Tone Reproduction Operator
TRC Tone Reproduction Curve
HDRI HDR image
RAW Raw HDR image
CONCLUSIONS
High dynamic images can be displayed with more details in
a low dynamic range display by using our method. The
resultant image contains average brightness on all parts with
more colour information. The number of exposures can be
from three to nineteen. Averaging pixel values with global
tone mapping is a fast and simple method of HDR
processing.
REFERENCES
This paper we used Java Netbeans for doing the operation
Tonemapping operators are applied on the images using the
above mentioned formula].We took {1},[2}, [3],[4], [5], [6]
as the references. We also took [7]. [8], [9] for the HDR
image and colour correction
[1] Jiang Duan and Guoping Qiu, “Fast Tone Mapping
for High Dynamic Range Images”

_______________________________________________________________________________________
[2] P. E. Debevec and J. Malik, “Recovering high dynamic
range radiance maps from photographs”, Proc. ACM
SIGGRAPH‟97, pp. 369 – 378, 1997
[3] J. DiCarlo and B. Wandell, “Rendering high dynamic
range images”, Proc. SPIE, vol.3965, pp. 392 – 401, 2001
[4] J. Tumblin and H. Rushmeier, “Tone reproduction for
realistic images”, IEEE Computer Graphics and
Applications, vol. 13, pp. 42 – 48, 1993
[5] M. Ashikhmin, “A tone mapping algorithm for high
contrast images”, Proc. Eurographics Workshop on
Rendering, P.Debevec and S. Gibson Eds., pp. 1 – 11, 2002
[6] G. W. Larson, H. Rushmeier and C. Piatko, “A visibility
matching tone reproduction operator for high dynamic range
scenes”, IEEE Trans on Visualization and Computer
Graphics, vol. 3, pp. 291 – 306, 1997
[7] K. Chiu, M. Herf, P. Shirley, S. Swamy, C. Wang and K.
Zimmerman, “Spatially nonuniform scaling functions for
high contrast images”, Proc. graphics Interface‟93, pp. 245
– 253, 1997
[8] J. Tumblin and G. Turk, “LCIS: A boundary hierarchy
for detail preserving contrast reduction”, ACM SIGGRAPH
1997
[9] F. Durand and J. Dorsey, “Fast bilateral filtering for the
display of high-dynamic- range images”, Proc. CM
SIGGRAPH‟2002
BIOGRAPHIES
Sajeena A, currently working as
Assistant Professor in Electronics
and Communication Engg
department at TKM College of
Engg. Kollam, Kerala, India
Sunitha Beevi K, working as
Assistant Professor in Electrical and
Electronics Engg. Department at
TKM College of Engg. Kollam,
Kerala, India

An improved hdr image processing using fast global tone mapping

More Related Content

What's hot (19)

Similar to An improved hdr image processing using fast global tone mapping (20)

More from eSAT Journals (20)

Recently uploaded (20)

An improved hdr image processing using fast global tone mapping