A Novel Low Complexity Histogram Algorithm for High Performance Image Processing Applications

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 01 | Jan -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1398
A Novel Low Complexity Histogram Algorithm for High Performance
Image Processing Applications
Govind Bhai1 and Shweta Agrawal2
1Research scholar, 2Assistant Professor,
Dept. of Electronics and Comm., SRCEM Banmore, Morena, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The modern portable devices requires low
complexity image processing designs to achieve high
performance economically. The histogram is used in modern
camera to capture the image of desired brightness. To
achieve high performance low cost histogram, low
complexity VLSI architectures are developed. The existing
architecture are not efficient, therefore this paper presents
four new low complexity histogram generator algorithm.
The effectiveness of the proposed algorithms is shown over
the existing by implementing and simulating them. The
simulation results show that the proposed algorithms show
significant reduction in implementation complexity.
Key Words - Image Processing, Histogram Generator,
Integrated Circuits, VLSI designs.
1. INTRODUCTION
The modern portable devices are embedding several
multimedia applications. The image processing
applications on these portable devices demands huge
computations and large power/area [1]. Due to the limited
battery and small size of portable devices, low complexity
design is required for image processing applications [2],
[3]. Among the different task used in the image processing,
the histogram is one of frequently used operation and
determines the performance of the application [4], [5], [6].
Although the histogram is being used in several
applications, most of the applications employ a software
program to achieve it and very little efforts have been
given to design a hardware architecture for the histogram
generator. This paper presents the limitations of existing
software level approach by implementing a VLSI
architecture for of the histogram.
In this paper, various existing architectures and
algorithms for histogram generation are reviewed and
analyzed [7], [8], [9], [10]. Further, four new algorithms
namely prop1, prop2, prop3 and prop4 are developed by
approximating pixels of nearly same value to a single
value. In the prop1, the pixels differ by one are
approximated to one value while in prop2 the pixels differ
by max of 3 value are approximated to single value. This is
achieved by truncation least significant bits (LSBs). In the
proposed architectures, the number of LSB bits truncated
are 1-bit, 2-bit, 3-bit, and 4-bit in prop1, prop2, prop3, and
prop4 design respectively to reduce the implementation
complexity.
In order to evaluate the efficacy of the proposed designs,
the proposed and existing designs are implemented and
simulated on MATLAB and in Verilog. The histogram
generated by the different histogram generator
architectures are compared where the simulation results
show the histogram using proposed algorithm provide
same information and exhibits identical structure over the
conventional design. Further, design implemented in
Verilog are syntheses and post synthesis results shows
that proposed architectures consume small area over the
conventional histogram generator. The area in terms of
LUTs register and LUT-BUF pairs is evaluated and
compared. Finally, the delay metrics is also computed and
compared for the different histogram architectures where
the proposed architectures show small delay.
The remainder of this paper explores different histogram
generator with comparative analysis by implementation
and simulation.
2. HISTOGRAM GENERATOR
An image histogram is a type of histogram that acts as a
graphical representation of the tonal distribution in a
digital image [11], [12]. A histogram of a set of data is
shown in Fig. 1 where each bar representing the frequency
of input data range.
Fig-1: An illustration of histogram for data range.

The histogram provides an easy way to perceive the
information regarding the frequency of sample data [13].
The histogram shows how spread/narrow the data is and
whether the data is located at the proper position or not as
shown in Fig. 2. It also shows whether the data is skewed
toward one direction or not. Finally, some common shapes
of histograms are shown in Fig. 3.
Fig-2: Histogram data spread and location.
Fig-3: Common shapes of the Histograms.
3. LOW COMPLEX HISTOGRAM ALGORITHS
This section present various algorithms and architectures
for low complexity histogram generation.
3.1 Programmable VLSI architecture for histogram
A programmable VLSI architecture for the histogram
generator is presented in [14]. This architecture is able to
compute the histograms of four free position-able window
which may overlap. This architecture computes the energy,
mean, maximum, minimum of the pixels which are having
the value above the given threshold. These value finally
determines the shape of the histogram. The histogram as
presented in this paper accumulates the number of pixels
of the same intensity.
3.2 Reconfigurable histogram architecture
A reconfigurable histogram architecture for the FPGA
based smart cameras is proposed by Maggiani et al. in [15].
The proposed histogram generator provides low
complexity of implementation and is suitable for the FPGA
based designs. The proposed histogram also does not
exhibit memory access conflict and employ very efficient
way of parallelism. The proposed architecture can provide
the histogram of input data i.e. employ the paradigm of
streaming. This architecture provides improved
performance over to the existing architectures. The
histogram core is the prime histogram generator block and
exhibits all steps of the histogram algorithm. This
architecture is reconfigurable and can provide the other
functionality. The prime advantage of this reconfigurable
architecture is the ability to work with real-time input and
providing full compatibility with the existing
reconfigurable architecture.
3.3 Low Complexity histogram architecture
A programmable histogram generator that can be
efficiently utilized in the image processing applications is
presented in Hazra et al. [16], [17]. The proposed
histogram generator architecture consists of two part
where one part is responsible for generating the histogram
data while other is responsible for displaying the results.
The architecture exhibits ‘operation selector block’ which
chooses kind of operation i.e. either generating the
histogram or displaying. This architecture also contains
ROM, decoder and array of counters. The ROM size is
216x8 used to hold the input image pixels while the
number of counters are 256 as there are 256 different
intensity levels. The different steps of the proposed
architectures are illustrated in Fig. 4.
Fig-4: Various steps of histogram generator.
Since the existing algorithms and architectures for
histogram still exhibit large implementation complexity,
new low complexity histogram architectures are proposed
and are presented in the following section.

4. PROPOSED LOW COMPLEX ALGORITHMS
The number of the registers depends on the different
intensity level considered in the histogram. Since in the
grayscale image there are 256 different value, it requires
256 memory registers to store the frequency of each
intensity. Further, a counter is used to increase the value of
each intensity once a pixel is encountered of that intensity.
These counters and registers requires very large space on
the IC or large number reconfigurable cells in the FPGA. In
a histogram, it is observed that the histogram will provide
the even if the pixels of values differed by 1 are combined
i.e. the pixels which are having value with difference of one.
By approximating two pixels in an image with values differ
by one, it reduces the number of different pixel intensity
from 256 to 128. Further, the truncation of two, three, and
four least significant bits in the pixels results in 64, 32 and
32 different intensity level only. In the proposed
algorithms the number of counters and registers are
reduced significantly. The proposed algorithms are shown
in shown below.
Algorithm1: Prop1
Input: image;
Output: histogram;
Initialize reg_array[1:128];
Initialize i, j, x;
[row, col] = size(image);
for i= 1: row
for j = 1: col
x= floor(image(i, j)/2);
increase reg_array(x);
end
end
Algorithm2: Prop2
Input: image;
Output: histogram;
Initialize i, j, x;
for i= 1: row
for j = 1: col
end
end
Algorithm3: Prop3 Algorithm4: Prop4
Input: image;
Output: histogram;
Initialize i, j, x;
for i= 1: row
for j = 1: col
end
end
Input: image;
Output: histogram;
Initialize i, j, x;
for i= 1: row
for j = 1: col
end
end
The proposed algorithms prop1, prop2, prop3, and
prop4 reduce the number of counters by 50%, 75%,
86.71%, and 93.75% respectively. The reduction in the
number of counters and registers reduces the power and
delay significantly. The histogram of the resulting image
will provide the same information as that of original while
significantly reducing the number of counters and
registers. Therefore, the proposed algorithms can be
effectively applied to reduce the implementation
complexity.
5. SIMULATION RESULT & ANALYSIS
To evaluate the efficacy of the proposed designs, all the
algorithms are implemented on MATLAB and simulated
with benchmark image (Lena). The histograms of the
image are extracted via proposed and existing
architectures. Further to achieve hardware analysis, the
algorithms are implemented in Verilog and area, delay
metrics are extracted.
The histogram of the Lena image using MATLAB inbuilt
function (imhist), conventional algorithm and proposed
algorithms are shown in Fig. 5 below. It can be observed
from these histograms that that the proposed algorithm
exhibits nearly same algorithm while conveying same
information.

(a) imhist (b) conventional
(c) prop1 (d)prop2
(e) prop3 (f) prop4
Fig-5: Histograms of Lena image using various algorithms.
On the other hand, to extract the hardware
analysis, all the designs are implemented in Verilog and
synthesized on FPGA using Xilinx ISE CAD tool [18]. The
functionality of the designs is verified. The RTL schematic
from the tool for the proposed algorithm prop2 is shown
in Fig. 6.
(a)
(b)
Fig-6: Proposed histograms generator a) Block diagram,
b) RTL schematic.
The post-synthesis results of the different histogram
generator are tabulated in Table 5.2. The area of the
different histogram architectures is computed in terms of
number of slice registers, slice-LUT and LUT-BUFF pairs.
Table-1: Design metrics of different histogram generator
architectures.
Metrics Conv. Prop1 Prop2 Prop3 Prop4
Delay (nS) 3.585 2.732 2.472 2.13 2.124
Frequency
(MHz) 278.9 365.9 404.7 469.4 470.7
#Slice Regs 8 7 6 5 4
# Slice LUTs 41 39 14 9 8
#LUT-BUF pair 8 7 6 5 0
IOBs 17 17 17 17 17
It can be observed from the Table 1 that the area
required by the proposed histogram algorithms are much
smaller over the conventional histogram architecture. The
proposed designs prop1, prop2, prop3 and prop4 reduce
the delay by 23.8%, 31.04%, 40.58%, and 40.86%
respectively. The comparison of the delay for various
histogram architectures are shown in Fig. 7.

Fig-7: Delay comparison for various histogram
architectures.
Further the operating frequency of the design
depends on the critical path delay, higher the value of the
frequency is required for the higher performance designs.
The operation frequency of the implemented designs is
also computed and compared as shown in Fig. 8 where it
can be seen that proposed designs provides higher
performance over the existing. Finally, the performance
increase with more approximation in the pixel’s values.
Fig-8: Performance of various histogram architectures.
The number of LUT-BUF pairs required by the
different histogram architectures are compared in Fig. 9. It
can be observed from the figure that the proposed design
prop4 requires very small number of LUT-BUF pairs over
the other architectures.
Fig-9: # LUT-BUF pairs for different histograms.
From the simulation results it is observed that
proposed design requires very small implementation area
over the conventional design while providing histograms
that exhibit same information as that by conventional
histogram generator. Therefore, the proposed
architectures are very useful for the applications where
implementation complexity is prime challenge.
6. CONCLUSION
The limited battery lifetime and the small size
requirement of the modern portable devices requires low
complexity designs. Most of these portable devices
exhibits multimedia applications, area efficient designs for
various image enhancement techniques are of serious
concern. In this paper novel low complexity algorithms for
the histogram generator are presented. The
implementation complexity of the proposed algorithms
very small due to reduce number of registers and
counters. To evaluate the effectiveness of the different
designs, all the filters architectures are implemented in
MATLAB and Tanner. The designs on the MATLAB are
simulated with benchmark input images and
corresponding scaled image and quality metrics are
extracted. The simulation results show that the proposed
filters requires 12.5%, 25%, 37.5%, and 50% reduced area
over the conventional histogram architecture. Moreover,
the proposed designs prop1, prop2, prop3 and prop4
reduce the delay by 23.8%, 31.04%, 40.58%, and 40.86%
respectively.
REFERENCES
[1]. T. Arslan, 1.M. Moreno, and A.M. Grigoryan, "New
Methods of Image Enhancement," in the Proceedings
of the SPIE conference on Visual Information
Processing XIV, 2007.

[2]. R.C. Gonzalez and R.E. Woods, Digital Image
Processing, Prentice Hall, 2nd Edition, 2001.
[3]. A. McAndrew, An Introduction to Digital Image
Processing with MATLAB Notes for SCM2511 Image
Processing 1, School of Computer Science and
Mathematics Victoria University of Technology, 2004.
[4]. Lorraine Denby and Colin Mallows, "Variations on the
histogram", Journal of Computational and Graphical
Statistics, January 2007.
[5]. P. Skelly, S. Dixit. and M. Schwanz, "A histogram-
based model for video traffic behaviour in an ATM
network node with an application to congestion
control." in Proceeding IEEE INFOCOM '92. Florence.
Italy. 1992.pp. 95-104.
[6]. Huaifeng Zhang, Xiangjian He, Qiang Wu, “Refined
Gaussian Weighted Histogram Intersection and Its
Application in Number Plate Categorization,” CGIV
2006, pp.249-255.
[7]. Muller, Steffen. "A new programmable VLSI
architecture for histogram and statistics computation
in different windows." Image Processing, 1995.
Proceedings., International Conference on. Vol. 1.
IEEE, 1995.
[8]. Cadenas, J., R. Simon Sherratt, and P. Huerta. "Parallel
pipelined histogram architectures." Electronics
letters 47.20 (2011): 1118-1120.
[9]. Gan, Q., J. M. P. Langlois, and Y. Savaria. "Parallel array
histogram architecture for embedded
implementations." Electronics Letters 49.2 (2013):
99-101.
[10].Shahbahrami, A., Hur, J.Y., Juulink, B., and Wong, S.:
‘FPGA implementation of parallel histogram
computation’. 2nd HiPEAC Workshop on
Reconfigurable Computing, Goteborg, Sweden, 2008,
pp. 63–72.
[11].Sanny A, Yang YH, Prasanna VK. Energy-efficient
histogram on FPGA. InReConFigurable Computing
and FPGAs (ReConFig), 2014 International
Conference on 2014 Dec 8 (pp. 1-6). IEEE.
[12].Hajinoroozi, M., Grigoryan, A., & Agaian, S. “Image
enhancement with weighted histogram equalization
and heap transforms”, In World Automation Congress
(WAC), 2016 (pp. 1-6). IEEE.
[13].K. S. Gautam, "Parallel Histogram Calculation for
FPGA: Histogram Calculation," 2016 IEEE 6th
International Conference on Advanced Computing
(IACC), Bhimavaram, 2016, pp. 774-777.
[14].Yin, S., Ouyang, P., Chen, T., Liu, L., & Wei, S. “A
Configurable Parallel Hardware Architecture for
Efficient Integral Histogram Image Computing”. IEEE
Transactions on Very Large Scale Integration (VLSI)
Systems, Vo. 24, No. 4, pp. 1305-1318.
[15].Maggiani, L., Salvadori, C., Petracca, M., Pagano, P., &
Saletti, R. (2014, June). Reconfigurable architecture
for computing histograms in real-time tailored to
FPGA-based smart camera. In Industrial Electronics
(ISIE), 2014 IEEE 23rd International Symposium on
(pp. 1042-1046). IEEE.
[16].Ghosh, S., Hazra, S., Maity, S. P., & Rahaman, H. “A new
algorithm for grayscale image histogram
computation”, In India Conference (INDICON), 2015
Annual IEEE (pp. 1-6). IEEE.
[17].Hazra, S., Ghosh, S., Maity, S. P., & Rahaman, H. “A New
FPGA and Programmable SoC Based VLSI
Architecture for Histogram Generation of Grayscale
Images for Image Processing Applications”, Procedia
Computer Science, Vol. 93, pp. 139-145.
ModelSim, “ModelSim HDL simulator,”
http://www.mentor.com/products/fpga/model, 2013.

A Novel Low Complexity Histogram Algorithm for High Performance Image Processing Applications

Recommended

More Related Content

What's hot (19)

Similar to A Novel Low Complexity Histogram Algorithm for High Performance Image Processing Applications (20)

More from IRJET Journal (20)

Recently uploaded (20)

A Novel Low Complexity Histogram Algorithm for High Performance Image Processing Applications