The document describes a decision tree based technique for removing impulse noise from digital images. It uses a 3x3 pixel mask to detect noisy pixels and then employs an edge-preserving filter to reconstruct pixel values. The technique was implemented on an FPGA and tested on test images corrupted with random valued impulse noise. It achieved better noise removal compared to other lower complexity methods while preserving image details due to its accurate noise detection and minimal hardware requirements.

1. Electronics & Communication
Engineering
(2015-16)

2. SAMIP KUNDU
RAJATH GOWDA
SHRIYA H.M.
SHIVANI PUSHPARAJ
Ms. USHA K.P.
Asst. Professor
B.E., M.Tech.
Department of ECE
AIT, Chikmagalur

3. Contents Abstract
 Introduction
 Objective
 Literature Survey
 Problem Definition & Formulation
 Decision Tree Based De-noising Method
 VLSI Implementation of DTBDM
 Specifications of Tools
 Overview & Applications of DTBDM
 Advantages & Future Enhancements
 Implementation Results
 Conclusion
 Bibliography

4. Abstract
Digital Image Processing is a promising area of research in the fields of
electronics and communication engineering. In this project, an efficient
very large scale integration (VLSI) and field programmable gate array
(FPGA) based impulsive noise detection technique is presented. This
design uses a 3x3 mask on each pixel in the image in order to determine
whether it is corrupted by random-valued impulse noise or not. We
employ a decision-tree-based impulse noise detector to detect the noise
pixels. After noise detection, the algorithm reconstructs the noisy pixel
by considering the possible edges existing in the mask. Due to its lower
complexity, the proposed technique is very suitable for hardware
implementation.

5. Introduction
 Digital Image Processing is a promising area of research in the
fields of electronics and communication engineering, consumer
and entertainment electronics, control and instrumentation,
biomedical instrumentation, remote sensing, robotics and
computer vision and computer aided manufacturing. For a
meaningful and useful processing such as image segmentation
and object recognition, and to have very good visual display in
applications like television, photo-phone, mobiles, etc…
 An image gets corrupted with noise during the processes of
acquisition, transmission, storage and retrieval. The digital
images are often corrupted by impulse noise due to transmission
errors, malfunctioning pixel elements in the camera sensors,
faulty memory locations, and timing errors in analog-to-digital
conversion.

6.  Impulse noise can be classified into two types: fixed-valued
impulse noise and random-valued impulse noise. The fixed-
valued impulse noise is also called salt-and-pepper noise
where the grey-scale value of a noisy pixel is either
minimum or maximum in grey-scale images. When viewed,
the image contains dark and white dots, hence the term salt
and pepper noise.
 In most applications, de-noising the image is fundamental
to subsequent image processing operations, such as edge
detection, image segmentation, object recognition, etc. The
goal of noise removal is to suppress the noise while
preserving image details.

7. Objective
 In this project, we propose an efficient denoising scheme and its VLSI
architecture for the removal of random-valued impulse noise.
 Our goal is to suppress noise, while preserving the image details.
 Our extensive experimental results demonstrate that the proposed
technique can obtain better performances in terms of both quantitative
evaluation and visual quality than the previous lower complexity methods.
 The design requires only low computational complexity and two line
memory buffers.
 This design implements minimal hardware, so hardware cost is low.
 We are trying to get a better reconstructed image as output, so that its
suitable to be applied to many real-time applications.

8. Literature Survey
 Many researchers have worked on impulse noise removal techniques,
like- median filter, ACWN, LCNR, RORD, DRID etc….
 Median filter removes the impulse noise keeping edges of the images
unaffected.
 ACWM filter works on switching method. A difference between output of
centre weighted median filter and the current pixel is calculated. With
this calculation a more general operator that depends upon impulse
detection is estimated.
 LCNR is implemented with two steps, noise detector and filtering. It
detects random valued noisy pixels and applies median filter only for
noisy pixels.
 RORD improves the impulse noise detection accuracy by using a
reference image. Then we introduce a simple weighted mean filter to
suppress the impulse noise while preserving image details.

9. Comparison of Different Techniques
Sr. no. Technique Complexity Advantages Disadvantages
1. Median Low Simple For Fixed
Impulse Noise
2. Adaptive Centre
Weighted Median
Low Suppresses both noise Reconstructed
Image is blur
3. Low Complexity
Noise Removal
Low Less logic elements
are used
Reconstructed
Image is blur
4. Adaptive Median
Filter
Low Good where fast
processing is required
Reconstructed
Image is blur
5. Alpha Trimmed Mean High De-noised image
quality is good
Requires full
frame buffer
6. Differential Rank
Impulse Detector
High De-noised image
quality is good
Requires four
iteration time
7. Rank ordered
Relative Difference
High High performance 7X7 mask size
is used
8. Decision Tree Based
De-noising (DTBDM)
Low All the Above
……

10. Problem Definition &
Formulation
 Images are often corrupted by impulse noise in the procedures of image
acquisition and transmission. Most filtering techniques work well with fixed
valued impulse noise. But today’s real time applications demand for an
efficient technique that can suppress both fixed and random valued impulse
noise.
 Nowadays, a good low complexity de-noising technique is necessary as pre-
processing operation in many real-time practical applications. In the process
of impulse noise filtering it is necessary to preserve edges and details of the
image. Also to avoid image smoothing, only corrupted pixel must be filtered.
 The most effective technique to remove random valued impulse noise without
losing useful information with pleasing denoised image is by decision-tree
based impulse detector and direction oriented edge preserving image filter.

11. Decision Tree Based De-noising Method
(DTBDM)
The decision tree is a simple but powerful form of multiple variable
analysis. It can break down a complex decision-making process into a
collection of simpler decisions, thus provide a solution which is often easier
to interpret .
WORKING
PRINCIPLE

12. EDGES
EDGES
Code: do_canny.m
canny_im = edge(image,‘canny');

13. Difference Between Noisy and Neighbouring Pixels
The Edge Region

14. Decision Tree Based Impulse Noise
Detector
 Observing the degree of isolation at current pixel.
 Determining whether the current pixel is on a fringe or
comparing similarity between current pixel and its
neighbouring pixels.
1. Isolation Module
2. Fringe Module
3. Similarity Module

15.  We determine whether current pixel is an isolation point by
observing the smoothness of its surrounding pixels.
 Finally we make a temporary decision whether pi,j is a
suspended noisy pixel or noise free.

16.  If Pi,j has a great difference with neighbouring pixels, it might be a noisy
pixel or just situated on the edge.
 We take E1 for example. By calculating absolute difference between fi,j
and other two pixels, we can determine its edge or not.

17.  The luminance values in mask W located in a noisy-free area might be
close.
 The median is always located in the centre of the variational series,
while the impulse is usually located near one of its ends. Hence, if there
are extreme big or small values, that implies the possibility of noisy
signals.
 If fi,j is not between Maxi,j and Mini,j, we conclude that pi,j is a noise
pixel. Edge-preserving image filter will be used to build the
reconstructed value. Otherwise, the original value fi,j will be the output.

18. VLSI Implementation of DTBDM
 The noise is generated by MATLAB function (0.3*randn(128))
 Amplitude of noise can be given as- (noise=0.3)

19. Specifications of Tools
SOFTWARE
 MATLAB
 Xilinx Platform Studio
HARDWARE
 Spartan 3 EDK Board
(Spartan 3 FPGA)

20. SOFTWARE
MATLAB
 MATLAB is a powerful language for technical computing. The name
MATLAB Stands for MATrix LABoratory, because its basic data
element is a matrix (array). MATLAB can be used for math
computations, modelling and simulations, data analysis and processing,
visualizations and graphics and algorithm development.
 MATLAB includes tools that allow a programmer to interactively
construct a GUI for his or her program. With this capability, the
programmer can design sophisticated data-analysis programs that can
be operated by relatively inexperienced users.

21. XILINX PLATFORM STUDIO
 Xilinx platform studio is a key component of the ISE embedded edition
design suite, helping the hardware designer to easily built, connect and
configure embedded processor based systems; from simple state
machines to full blown 32-bit RISC microprocessor systems.
 XPS employs graphical design views and sophisticated correct by
design wizard to guide developers though the steps necessary to create
custom processor system within minutes.
 The true potential of XPS emerges with its ability to configure and
integrate plug and play IP cores from the Xilinx embedded IP
catalogue, with custom or third party Verilog and VHDL designs.

22. HARDWARE

26. Overview & Applications
of DTBDM
 Here, we have created the noisy image in MATLAB and the hex value of the
image (image.h file) is stored in SDRAM using RS232 serial port.
 This is further read into the SRAM (input buffer) and is sent for processing
through Direct Memory Access (DMA) and the filtered image is stored in
SRAM (output buffer).
 Finally, we observe the restored image in Visual Basic (VB) window.
 Image processing is widely used in many fields, such as medical imaging,
scanning techniques, printing skills, license plate recognition, face
recognition, and so on.
 The noise may seriously affect the performance of image processing
techniques. Hence, in such situations DTBDM technique is very necessary.

27. Advantages
 Suppresses both fixed and random valued impulse noise
 Uses only 3x3 mask
 Uses only 2 line buffer and less memory
 Low complexity
 Low cost
Future Enhancements
This technique can be used for real time applications like scanning, face-
recognition, edge detection, medical imaging, printing, license plate
detection, where it is important to remove noises before these subsequent
processes. DTBDM technique can be further used in future for video
processing in televisions, mobiles, computers, gaming with high graphics
etc.

28. Implementation Results
 To verify the characteristics and performances of DTBDM,
it is implemented on 128x128 8-bit gray scale test image.
Original Image Noisy Image Restored Image

29. Original Image Sending
Received Restored Image

30. Conclusion
 In this project, we have presented an efficient decision-based filter for
noise detection and image restoration.
 Because the new impulse detection mechanism can accurately tell where
noise is, only the noise-corrupted pixels are replaced with the estimated
central noise-free ordered mean value.
 As a result, the restored images can preserve perceptual details and edges
in the image while effectively suppressing impulse noise.
 The VLSI architecture of our design requires only low computational
complexity and two line memory buffers hence making it suitable for
real-time applications.
 The architectures work with monochromatic images, but they can be
extended for working with RGB color images and videos.

31. Bibliography
 R.C. Gonzalez and R.E. Woods, Digital Image Processing,
Pearson Education, New Jersey, 2007.
 W.K. Pratt, Digital Image Processing, New York: Wiley-
Inter-science, 1991.
 P.-Y. Chen and C.-Y. Lien, “An Efficient Edge-Preserving
Algorithm for Removal of Salt-and-Pepper Noise,”
IEEE Signal Process. Dec.2008.
 A.S. Awad and H. Man, “High performance detection filter
for impulse noise removal in images,” IEEE Electron, Jan
2008.