SlideShare a Scribd company logo
1 of 26
TERM PAPER ON
EFFICIENT BLOCK MATCHING FOR
REMOVING IMPULSE NOISE
GUIDED BY:
Dr.M.Pompapathi
PRESENTED BY:
V.Yuvraj(Y20IT124)
T.Bharath kumar(Y20IT114)
Mohammad Afride(Y20IT077)
CONTENTS
1 . ABSTRACT
2 . KEYWORDS
3 . FLOW CHART
4 . ALGORITHM FOR PROPOSED WORK
5 . PROCESS OF THE ALGORITHM
6 . EXPERIMENTAL RESULTS
7 . CONCLUSION
8 . REFERENCES
ABSTRACT
A number of block-based image-denoising methodshave been presented in
the literature. Those methods, however,are generally adapted to denoising
the Gaussian noise, and sub sequently do not show good performance for
denoising random valued impulse, and salt-and-pepper noise. We propose
an efficient block-based image-denoising method, which is devised
specially forfast denoising of impulse noise. The method first constructs a
setof array pointers to image blocks containing a specific pixel valueat a
specific location. With this scheme, finding of blocks similar to a given
block can be done by considering only the blockspointed by the pointers
corresponding to the pixel values of theblock without comparing all the
blocks in the input image. The experimental results show that the proposed
method can achievesuperior denoising performance in terms of
computational timeand signal-to-noise ratio measure.
KEYWORDS
1 .IMAGE DENOISING : Computer vision task that involves removing
noise from an image.
2 .SALT & PEPPER NOISE : Form of noise that can be seen on images.
3 .MEDIAN FILTER : Non-linear digital filtering technique , often used to
remove noise from images or signal.
4 .BMLUT ALGORITHM : Block matching by look up table.
5 .EUCLEDIAN DISTANCE : The straight-line distance between two
pixels. The city block distance metric measures the path between the pixels
based on a 4-connected neighborhood. Pixels whose edges touch are 1 unit
apart; pixels diagonally touching are 2 units apart.
6 .TOLERENCE FUNCTION : The percentage of pixels that can be a
mismatch between the two images and the images are considered to still be a
match.
7 .DIRECTIONAL WEIGHTED MEDIAN FILTER : The directional
weighted median filter [DWMF] is used to calculate the sum of absolute
difference between the gray scale values in all the four directions of the
selected window.
8 .PEAK SIGNAL TO NOISE(PSNR) : An expression for the ratio between
the maximum possible value (power) of a signal and the power of distorting
noise that affects the quality of its representation.
9 . NOISY PIXELS : Pixels that are made up of the sum of the original
values plus the gaussian noise pixels.
FLOW CHART
BMLUT FLOW CHART
BMLUT ALGORITHM
Input: noisy image I, Output: denoised image J,
Set the window size: 3x3, Set the threshold T,
Set the sufficient number of matching blocks: #S
1 Scan I with a stride of one pixel to obtain a set of matching
blocks, BM= {M: the number of noise in M is less than 5}.
2 Scan I with a stride of 3 pixels to obtain a set of non-overlapping
target blocks, BT = { T: the number of noise in T' is less than 5},
and a set of non-overlapping noisy blocks BN = {N: N=I, N&BT}.
3 From BM, construct a 3-D array A[0..8][0..255][MAX_PIX]
4 For tolerance d increasing by Ad from ds to de do:
5 While the number of T's in B¹ decreases do:
6 For each T' in BT do:
7 For each clean pixel of value p at location X₁ do:
8 Search the lists pointed by A[i][p-d..p+d] for
matching blocks and move the blocks to Σ(Τ).
9 If the number of collected matching bocks is
greater than #S for each of all the noise in T, then
stop the search.
10 Compute the mean of clean pixel values for M's in
Σ(7) and replace the noise in Twith the means.
11 If all noise in T'is denoised, move T from BT to BM,
and write the denoised pixels to J
12 For each N in BN do:
Divide N into four blocks, and if noise in some blocks are
less than 5, register them as new target blocks, and do
the operations of step 5 through 8.
13 If BT is not empty, go to step 4.
14 If BT and BN are not empty, the blocks in BT and BN are divided
into 2x2 blocks containing only one noisy pixel, and block
matching is performed until both BT and BN are empty.
PROCESS OF THE ALGORITHM :
A. Identification of Noisy Pixels
We determine if a pixel is corrupted by the local homogeneity assumption
around a pixel [8]. First, we build a two-dimensional(2-D) histogram H of
size 256 × 256, where (i, j)th element H(i, j), is the total number of
neighboring pixels of value j aroundpixels of value i. Fig. 1 shows the shape
of H where the vertical cross section of H at some point i on the x-axis
denotes the distributions of pixel values adjacent to all the pixels of valuei.
For each pixel value i and a given percentage δ, we definethe range of δ-
homogeneity by two values: 1) lowδ (i); and 2)upδ (i), so that the sum of H(i,
j) with j running from lowδ (i) toupδ (i) equals to δ percent of the sum of all
the bins on the cross section at i .
After all of lowδ (i) and upδ (i) for all i from 0 to 255 have been computed ,
noisy pixels are determined as follows. For eachpixel q adjacent to p, if p is in
the range of q’s δ-homogeneity, then the counter is increased by 1. After all the
eight neigh-boring pixels are considered, if the counter is less than a threshold,
then p is class-ified as noise.
B. Construction of Pointer Arrays for Matching Blocks
Once all noisy pixels have been determined, a sliding windowW of size 3 × 3
moves over the input image I pixel by pixel, andnoisy pixels in W are counted.
If the number of noisy pixels in W is greater than a predefined threshold, it is
class-ified as a noisy block. Otherwise, the block is initially classified as a
matching block.
Among the matching blocks, the blocks including at least one noisy pixel are
tar-get blocks to be denoised using othersimilar matching blocks. In order to
reduce the computing time in searching similar blocks, we represent the pixel
locations in
a 3 × 3 window as numbers from 1 through 9, and maintain a set of 256-D
arrays of pointers {Ai[ 0 ... 255]: i = 1, 2,...,9} where each i corresponds to a
pixel loca-tion as shown in Fig. 2. Arrays {Ai} play a role of a lookup-table in
which the
information of “a pixel of value p is occurring at position i ina 3 × 3 block” is
recorded in the list pointed by Ai[p]. Thus, because Ai[p] points to the list
conta-ining the coordinates ofall the matching blocks whose ith pixel has value
of p, search of similar blocks can be done just by looking up the lists.
C. Noise Removal
Denoising of Target Blocks. Let Inoise denote the set of noisypixels in I.
Suppose the arrays Ai, i = 1, 2, . . . , 9, have beenconstructed as described earlier.
Next step is to collect similar matching blocks for each target block.
Computation of the distance between target and matching blocks involves noise,
and hence the Eucl-idian distance cannot be directly employed because
comparison of noise is mean-ingless. We propose a conceptof tolerance level d
Tolerance function determines whether two pixels are similar or not. Now,
supp-ose we are comparing a target block T = [t1 , t2 ,..., t9 ] where ti is a pixel
value at position i and amatching block M = [p1 , p2 , ..., p9 ] which is stored
in the lists pointed by Ai’s. We define the similarity of T and M as
In (3), noise cannot be involved in computing the similarity of two blocks. Our
goal is to build a set of matching blocks Σ(T) for T, which satisfy the following
condition for a threshold τ :
The search process continues until sufficient number #S, say atleast five, of
matching blocks are collected in Σ(T). If the number of matching blocks in Σ(T)
is not sufficient, tolerance level d is increased by Δd and the search process
starts again until sufficient numbers of matching blocks are collected.
Denoising of T is done by replacing the noisy pixels in T with the average of all
the clean pixels at the corresponding location of all the matching blocks in Σ(T).
On the left-hand side,target block T = [10, 15, 9, •, •, •, 8, •, 5] is given, where
“•” denotes noise. As x1 of T is 10 and d is 1, all the blocks pointed byA1 [ 10
− d..10 + d] = A1 [ 9..11] are candidates for matching to T. Among them, Fig. 3
shows that only M1 and M2 are matching to T, because simil(T, M1 ) = 4, and
simil(T, M2 ) =4, and hence M1 and M2 satisfy the condition of (5).
EXPERIMENTAL RESULTS
We evaluated the proposed BMLUT method using four images corrupted by
random-valued impulse noise of various rates. Fig. 6 shows the test images
along with 30% noisy images. In the first experiments, we assessed the gain of
comp-utational time of the proposed BMLUT method over the exhaustive
search_x0002_based method, and the running time of the directional weighted
median filter (DWMF) [9] was included just for comparing to nonblocking
matching method. Table I provides the average run time of each method on the
Intel Core i7-6700 3.40 GHz pro_x0002_cessor with 16 GB memory. The
tolerance level was set to from 3 to 9 increasing by Δd = 2. One can see that
the proposed BMLUT method greatly outperforms other methods. Fig. 7
showsthe intermediate denoising result of the Lena with 30% impulsenoise
after the iterations from steps 4 13 is over. Some noisypixels inside the circles
could not be denoised by 3 × 3 matching blocks within the tolerance less than
10, and therefore we need to divide the blocks containing the noisy pixels.
We also evaluated the denoising capability of the BMLUT method and
compared the results with other methods, includingthe standard median filter
using the peak signal to noise (PSNR) measure,
where MAXI refers to the max of the gray level which is 255 for 8-bit images
and MSE stands for mean-squared error between the input and the restored
images. Table II shows the quality of the denoised images for the four test
images. One can see that the proposed BMLUT method consistently
outperforms other methods, including the recently published three valued
weighted filter (TVWF) approach [5]. In addition to the objective measure
PSNR, we compared the subjective visual quality of the restored images as
illustrated in the Supplementary material.
CONCLUSION
We proposed an efficient block-matching method, which is superior for
denoising the input image corrupted by random valued impulse noise. In the
preprocessing stage, the noisy pixels are identified using the local
homogeneity property. With this information of locations of noisy pixels, the
proposed method divides blocks into matching and noisy blocks, based on the
number of noisy pixels in the block. Among matching blocks, nonoverlapping
blocks containing at least one noisy pixel are called target blocks and are
subject to denoising. The denoising operation is carried out using the array of
pointers, which allows to determine what pixel values occur at which
locations in the image, and hence reduces the search time to find the matching
blocks. Our method has shown superior performance gain in terms of
computing time and quality of the restored images.
REFERENCES
[1] N. Yao and K. K. Ma, “Adaptive rood pattern search for fast block-
matching
motion estimation”. IEEE Trans. Image Process., vol. 11 no. 12, pp. 1442–1448,
Dec. 2002.
[2] C. Je and H.-M. Park, “Optimized hierarchical block matching for fast and
acc-urate image registration,” Signal Process.: Image Commun., vol. 28, no. 7,
pp. 779–791, 2013.
[3] K. Dabov, A. Foi, and K. Egiazarian“Image denoising by sparse 3D
transform-domain collaborative filtering,” IEEE Trans. Image Process.,vol. 16,
no. 8, pp. 2080–2095, Aug. 2007.
[4] B. Ahn and N. I. Cho, “Block-matching convolutional neural network
forimage denoising,” CoRR abs/1704.00524, 2017.
[5] A. Roy, J. Singha, S. S. Devi, and R. H. Laskar, “Impulse noise removal
using SVM classification based fuzzy filter from gray scale images,” Signal
Process., vol. 28, pp. 262–273, 2016.
[6] A. Roy and R. H. Laskar, “Non-casual linear prediction based adaptive
filter for removal of high density impulse noise from color images,” AEU Int. J.
Electron. Commun. vol. 72, pp. 114–124, 2017.
[7] G. Pok, J. C. Liu, and A. S. Nair, “Selective removal of impulse noise
based on homogeneity level information,” IEEE Trans. Image Process., vol. 12,
no. 1, pp. 85–92, Jan. 2003.
[8] Y. Dong and S. Xu, “A New Directional weighted median filter for removal
of random-valued impulse noise,” IEEE Signal Process. Lett.,vol. 14, no. 3, pp.
193–196, Mar. 2007.
THANK YOU

More Related Content

Similar to Efficient block matching for removing impulse noise

IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD Editor
 

Similar to Efficient block matching for removing impulse noise (20)

Reduced Ordering Based Approach to Impulsive Noise Suppression in Color Images
Reduced Ordering Based Approach to Impulsive Noise Suppression in Color ImagesReduced Ordering Based Approach to Impulsive Noise Suppression in Color Images
Reduced Ordering Based Approach to Impulsive Noise Suppression in Color Images
 
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
 
Direction of arrival estimation using music algorithm
Direction of arrival estimation using music algorithmDirection of arrival estimation using music algorithm
Direction of arrival estimation using music algorithm
 
Direction of arrival estimation using music algorithm
Direction of arrival estimation using music algorithmDirection of arrival estimation using music algorithm
Direction of arrival estimation using music algorithm
 
K010615562
K010615562K010615562
K010615562
 
Edge Detection with Detail Preservation for RVIN Using Adaptive Threshold Fil...
Edge Detection with Detail Preservation for RVIN Using Adaptive Threshold Fil...Edge Detection with Detail Preservation for RVIN Using Adaptive Threshold Fil...
Edge Detection with Detail Preservation for RVIN Using Adaptive Threshold Fil...
 
Identification system of characters in vehicular plates
Identification system of characters in vehicular platesIdentification system of characters in vehicular plates
Identification system of characters in vehicular plates
 
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
 
Mm chap08 -_lossy_compression_algorithms
Mm chap08 -_lossy_compression_algorithmsMm chap08 -_lossy_compression_algorithms
Mm chap08 -_lossy_compression_algorithms
 
Removal of Fixed-valued Impulse Noise based on Probability of Existence of t...
Removal of Fixed-valued Impulse Noise based on Probability  of Existence of t...Removal of Fixed-valued Impulse Noise based on Probability  of Existence of t...
Removal of Fixed-valued Impulse Noise based on Probability of Existence of t...
 
www.ijerd.com
www.ijerd.comwww.ijerd.com
www.ijerd.com
 
THE EVIDENCE THEORY FOR COLOR SATELLITE IMAGE COMPRESSION
THE EVIDENCE THEORY FOR COLOR SATELLITE IMAGE COMPRESSIONTHE EVIDENCE THEORY FOR COLOR SATELLITE IMAGE COMPRESSION
THE EVIDENCE THEORY FOR COLOR SATELLITE IMAGE COMPRESSION
 
Algorithm
AlgorithmAlgorithm
Algorithm
 
Lossy
LossyLossy
Lossy
 
Bucket sort- A Noncomparision Algorithm
Bucket sort- A Noncomparision AlgorithmBucket sort- A Noncomparision Algorithm
Bucket sort- A Noncomparision Algorithm
 
3 ijaems nov-2015-6-development of an advanced technique for historical docum...
3 ijaems nov-2015-6-development of an advanced technique for historical docum...3 ijaems nov-2015-6-development of an advanced technique for historical docum...
3 ijaems nov-2015-6-development of an advanced technique for historical docum...
 
www.ijerd.com
www.ijerd.comwww.ijerd.com
www.ijerd.com
 
Real Time Speech Enhancement in the Waveform Domain
Real Time Speech Enhancement in the Waveform DomainReal Time Speech Enhancement in the Waveform Domain
Real Time Speech Enhancement in the Waveform Domain
 
dsp
dspdsp
dsp
 
INFLUENCE OF QUANTITY OF PRINCIPAL COMPONENT IN DISCRIMINATIVE FILTERING
INFLUENCE OF QUANTITY OF PRINCIPAL COMPONENT IN DISCRIMINATIVE FILTERINGINFLUENCE OF QUANTITY OF PRINCIPAL COMPONENT IN DISCRIMINATIVE FILTERING
INFLUENCE OF QUANTITY OF PRINCIPAL COMPONENT IN DISCRIMINATIVE FILTERING
 

Recently uploaded

一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
tuuww
 
ENCODERS & DECODERS - Digital Electronics - diu swe
ENCODERS & DECODERS - Digital Electronics - diu sweENCODERS & DECODERS - Digital Electronics - diu swe
ENCODERS & DECODERS - Digital Electronics - diu swe
MohammadAliNayeem
 
Online blood donation management system project.pdf
Online blood donation management system project.pdfOnline blood donation management system project.pdf
Online blood donation management system project.pdf
Kamal Acharya
 

Recently uploaded (20)

Attraction and Repulsion type Moving Iron Instruments.pptx
Attraction and Repulsion type Moving Iron Instruments.pptxAttraction and Repulsion type Moving Iron Instruments.pptx
Attraction and Repulsion type Moving Iron Instruments.pptx
 
Research Methodolgy & Intellectual Property Rights Series 2
Research Methodolgy & Intellectual Property Rights Series 2Research Methodolgy & Intellectual Property Rights Series 2
Research Methodolgy & Intellectual Property Rights Series 2
 
Roushan Kumar Java oracle certificate
Roushan Kumar Java oracle certificate Roushan Kumar Java oracle certificate
Roushan Kumar Java oracle certificate
 
一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
一比一原版(UNK毕业证)内布拉斯加州立大学科尼分校毕业证成绩单
 
Natalia Rutkowska - BIM School Course in Kraków
Natalia Rutkowska - BIM School Course in KrakówNatalia Rutkowska - BIM School Course in Kraków
Natalia Rutkowska - BIM School Course in Kraków
 
Supermarket billing system project report..pdf
Supermarket billing system project report..pdfSupermarket billing system project report..pdf
Supermarket billing system project report..pdf
 
ONLINE CAR SERVICING SYSTEM PROJECT REPORT.pdf
ONLINE CAR SERVICING SYSTEM PROJECT REPORT.pdfONLINE CAR SERVICING SYSTEM PROJECT REPORT.pdf
ONLINE CAR SERVICING SYSTEM PROJECT REPORT.pdf
 
An improvement in the safety of big data using blockchain technology
An improvement in the safety of big data using blockchain technologyAn improvement in the safety of big data using blockchain technology
An improvement in the safety of big data using blockchain technology
 
Research Methodolgy & Intellectual Property Rights Series 1
Research Methodolgy & Intellectual Property Rights Series 1Research Methodolgy & Intellectual Property Rights Series 1
Research Methodolgy & Intellectual Property Rights Series 1
 
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWINGBRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
 
İTÜ CAD and Reverse Engineering Workshop
İTÜ CAD and Reverse Engineering WorkshopİTÜ CAD and Reverse Engineering Workshop
İTÜ CAD and Reverse Engineering Workshop
 
Electrostatic field in a coaxial transmission line
Electrostatic field in a coaxial transmission lineElectrostatic field in a coaxial transmission line
Electrostatic field in a coaxial transmission line
 
ENCODERS & DECODERS - Digital Electronics - diu swe
ENCODERS & DECODERS - Digital Electronics - diu sweENCODERS & DECODERS - Digital Electronics - diu swe
ENCODERS & DECODERS - Digital Electronics - diu swe
 
Online resume builder management system project report.pdf
Online resume builder management system project report.pdfOnline resume builder management system project report.pdf
Online resume builder management system project report.pdf
 
Online blood donation management system project.pdf
Online blood donation management system project.pdfOnline blood donation management system project.pdf
Online blood donation management system project.pdf
 
Cloud-Computing_CSE311_Computer-Networking CSE GUB BD - Shahidul.pptx
Cloud-Computing_CSE311_Computer-Networking CSE GUB BD - Shahidul.pptxCloud-Computing_CSE311_Computer-Networking CSE GUB BD - Shahidul.pptx
Cloud-Computing_CSE311_Computer-Networking CSE GUB BD - Shahidul.pptx
 
Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...
 
2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edge2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edge
 
Arduino based vehicle speed tracker project
Arduino based vehicle speed tracker projectArduino based vehicle speed tracker project
Arduino based vehicle speed tracker project
 
Dairy management system project report..pdf
Dairy management system project report..pdfDairy management system project report..pdf
Dairy management system project report..pdf
 

Efficient block matching for removing impulse noise

  • 1. TERM PAPER ON EFFICIENT BLOCK MATCHING FOR REMOVING IMPULSE NOISE GUIDED BY: Dr.M.Pompapathi PRESENTED BY: V.Yuvraj(Y20IT124) T.Bharath kumar(Y20IT114) Mohammad Afride(Y20IT077)
  • 2. CONTENTS 1 . ABSTRACT 2 . KEYWORDS 3 . FLOW CHART 4 . ALGORITHM FOR PROPOSED WORK 5 . PROCESS OF THE ALGORITHM 6 . EXPERIMENTAL RESULTS 7 . CONCLUSION 8 . REFERENCES
  • 3. ABSTRACT A number of block-based image-denoising methodshave been presented in the literature. Those methods, however,are generally adapted to denoising the Gaussian noise, and sub sequently do not show good performance for denoising random valued impulse, and salt-and-pepper noise. We propose an efficient block-based image-denoising method, which is devised specially forfast denoising of impulse noise. The method first constructs a setof array pointers to image blocks containing a specific pixel valueat a specific location. With this scheme, finding of blocks similar to a given block can be done by considering only the blockspointed by the pointers corresponding to the pixel values of theblock without comparing all the blocks in the input image. The experimental results show that the proposed method can achievesuperior denoising performance in terms of computational timeand signal-to-noise ratio measure.
  • 4. KEYWORDS 1 .IMAGE DENOISING : Computer vision task that involves removing noise from an image. 2 .SALT & PEPPER NOISE : Form of noise that can be seen on images. 3 .MEDIAN FILTER : Non-linear digital filtering technique , often used to remove noise from images or signal. 4 .BMLUT ALGORITHM : Block matching by look up table. 5 .EUCLEDIAN DISTANCE : The straight-line distance between two pixels. The city block distance metric measures the path between the pixels based on a 4-connected neighborhood. Pixels whose edges touch are 1 unit apart; pixels diagonally touching are 2 units apart.
  • 5. 6 .TOLERENCE FUNCTION : The percentage of pixels that can be a mismatch between the two images and the images are considered to still be a match. 7 .DIRECTIONAL WEIGHTED MEDIAN FILTER : The directional weighted median filter [DWMF] is used to calculate the sum of absolute difference between the gray scale values in all the four directions of the selected window. 8 .PEAK SIGNAL TO NOISE(PSNR) : An expression for the ratio between the maximum possible value (power) of a signal and the power of distorting noise that affects the quality of its representation.
  • 6. 9 . NOISY PIXELS : Pixels that are made up of the sum of the original values plus the gaussian noise pixels.
  • 9. BMLUT ALGORITHM Input: noisy image I, Output: denoised image J, Set the window size: 3x3, Set the threshold T, Set the sufficient number of matching blocks: #S 1 Scan I with a stride of one pixel to obtain a set of matching blocks, BM= {M: the number of noise in M is less than 5}. 2 Scan I with a stride of 3 pixels to obtain a set of non-overlapping target blocks, BT = { T: the number of noise in T' is less than 5}, and a set of non-overlapping noisy blocks BN = {N: N=I, N&BT}. 3 From BM, construct a 3-D array A[0..8][0..255][MAX_PIX] 4 For tolerance d increasing by Ad from ds to de do: 5 While the number of T's in B¹ decreases do:
  • 10. 6 For each T' in BT do: 7 For each clean pixel of value p at location X₁ do: 8 Search the lists pointed by A[i][p-d..p+d] for matching blocks and move the blocks to Σ(Τ). 9 If the number of collected matching bocks is greater than #S for each of all the noise in T, then stop the search. 10 Compute the mean of clean pixel values for M's in Σ(7) and replace the noise in Twith the means. 11 If all noise in T'is denoised, move T from BT to BM, and write the denoised pixels to J
  • 11. 12 For each N in BN do: Divide N into four blocks, and if noise in some blocks are less than 5, register them as new target blocks, and do the operations of step 5 through 8. 13 If BT is not empty, go to step 4. 14 If BT and BN are not empty, the blocks in BT and BN are divided into 2x2 blocks containing only one noisy pixel, and block matching is performed until both BT and BN are empty.
  • 12. PROCESS OF THE ALGORITHM : A. Identification of Noisy Pixels We determine if a pixel is corrupted by the local homogeneity assumption around a pixel [8]. First, we build a two-dimensional(2-D) histogram H of size 256 × 256, where (i, j)th element H(i, j), is the total number of neighboring pixels of value j aroundpixels of value i. Fig. 1 shows the shape of H where the vertical cross section of H at some point i on the x-axis denotes the distributions of pixel values adjacent to all the pixels of valuei. For each pixel value i and a given percentage δ, we definethe range of δ- homogeneity by two values: 1) lowδ (i); and 2)upδ (i), so that the sum of H(i, j) with j running from lowδ (i) toupδ (i) equals to δ percent of the sum of all the bins on the cross section at i .
  • 13. After all of lowδ (i) and upδ (i) for all i from 0 to 255 have been computed , noisy pixels are determined as follows. For eachpixel q adjacent to p, if p is in the range of q’s δ-homogeneity, then the counter is increased by 1. After all the eight neigh-boring pixels are considered, if the counter is less than a threshold, then p is class-ified as noise. B. Construction of Pointer Arrays for Matching Blocks Once all noisy pixels have been determined, a sliding windowW of size 3 × 3 moves over the input image I pixel by pixel, andnoisy pixels in W are counted. If the number of noisy pixels in W is greater than a predefined threshold, it is class-ified as a noisy block. Otherwise, the block is initially classified as a matching block.
  • 14. Among the matching blocks, the blocks including at least one noisy pixel are tar-get blocks to be denoised using othersimilar matching blocks. In order to reduce the computing time in searching similar blocks, we represent the pixel locations in a 3 × 3 window as numbers from 1 through 9, and maintain a set of 256-D arrays of pointers {Ai[ 0 ... 255]: i = 1, 2,...,9} where each i corresponds to a pixel loca-tion as shown in Fig. 2. Arrays {Ai} play a role of a lookup-table in which the information of “a pixel of value p is occurring at position i ina 3 × 3 block” is recorded in the list pointed by Ai[p]. Thus, because Ai[p] points to the list conta-ining the coordinates ofall the matching blocks whose ith pixel has value of p, search of similar blocks can be done just by looking up the lists.
  • 15. C. Noise Removal Denoising of Target Blocks. Let Inoise denote the set of noisypixels in I. Suppose the arrays Ai, i = 1, 2, . . . , 9, have beenconstructed as described earlier. Next step is to collect similar matching blocks for each target block. Computation of the distance between target and matching blocks involves noise, and hence the Eucl-idian distance cannot be directly employed because comparison of noise is mean-ingless. We propose a conceptof tolerance level d
  • 16. Tolerance function determines whether two pixels are similar or not. Now, supp-ose we are comparing a target block T = [t1 , t2 ,..., t9 ] where ti is a pixel value at position i and amatching block M = [p1 , p2 , ..., p9 ] which is stored in the lists pointed by Ai’s. We define the similarity of T and M as In (3), noise cannot be involved in computing the similarity of two blocks. Our goal is to build a set of matching blocks Σ(T) for T, which satisfy the following condition for a threshold τ :
  • 17. The search process continues until sufficient number #S, say atleast five, of matching blocks are collected in Σ(T). If the number of matching blocks in Σ(T) is not sufficient, tolerance level d is increased by Δd and the search process starts again until sufficient numbers of matching blocks are collected. Denoising of T is done by replacing the noisy pixels in T with the average of all the clean pixels at the corresponding location of all the matching blocks in Σ(T).
  • 18. On the left-hand side,target block T = [10, 15, 9, •, •, •, 8, •, 5] is given, where “•” denotes noise. As x1 of T is 10 and d is 1, all the blocks pointed byA1 [ 10 − d..10 + d] = A1 [ 9..11] are candidates for matching to T. Among them, Fig. 3 shows that only M1 and M2 are matching to T, because simil(T, M1 ) = 4, and simil(T, M2 ) =4, and hence M1 and M2 satisfy the condition of (5).
  • 19. EXPERIMENTAL RESULTS We evaluated the proposed BMLUT method using four images corrupted by random-valued impulse noise of various rates. Fig. 6 shows the test images along with 30% noisy images. In the first experiments, we assessed the gain of comp-utational time of the proposed BMLUT method over the exhaustive search_x0002_based method, and the running time of the directional weighted median filter (DWMF) [9] was included just for comparing to nonblocking matching method. Table I provides the average run time of each method on the Intel Core i7-6700 3.40 GHz pro_x0002_cessor with 16 GB memory. The tolerance level was set to from 3 to 9 increasing by Δd = 2. One can see that the proposed BMLUT method greatly outperforms other methods. Fig. 7 showsthe intermediate denoising result of the Lena with 30% impulsenoise after the iterations from steps 4 13 is over. Some noisypixels inside the circles could not be denoised by 3 × 3 matching blocks within the tolerance less than 10, and therefore we need to divide the blocks containing the noisy pixels.
  • 20. We also evaluated the denoising capability of the BMLUT method and compared the results with other methods, includingthe standard median filter using the peak signal to noise (PSNR) measure,
  • 21.
  • 22. where MAXI refers to the max of the gray level which is 255 for 8-bit images and MSE stands for mean-squared error between the input and the restored images. Table II shows the quality of the denoised images for the four test images. One can see that the proposed BMLUT method consistently outperforms other methods, including the recently published three valued weighted filter (TVWF) approach [5]. In addition to the objective measure PSNR, we compared the subjective visual quality of the restored images as illustrated in the Supplementary material.
  • 23. CONCLUSION We proposed an efficient block-matching method, which is superior for denoising the input image corrupted by random valued impulse noise. In the preprocessing stage, the noisy pixels are identified using the local homogeneity property. With this information of locations of noisy pixels, the proposed method divides blocks into matching and noisy blocks, based on the number of noisy pixels in the block. Among matching blocks, nonoverlapping blocks containing at least one noisy pixel are called target blocks and are subject to denoising. The denoising operation is carried out using the array of pointers, which allows to determine what pixel values occur at which locations in the image, and hence reduces the search time to find the matching blocks. Our method has shown superior performance gain in terms of computing time and quality of the restored images.
  • 24. REFERENCES [1] N. Yao and K. K. Ma, “Adaptive rood pattern search for fast block- matching motion estimation”. IEEE Trans. Image Process., vol. 11 no. 12, pp. 1442–1448, Dec. 2002. [2] C. Je and H.-M. Park, “Optimized hierarchical block matching for fast and acc-urate image registration,” Signal Process.: Image Commun., vol. 28, no. 7, pp. 779–791, 2013. [3] K. Dabov, A. Foi, and K. Egiazarian“Image denoising by sparse 3D transform-domain collaborative filtering,” IEEE Trans. Image Process.,vol. 16, no. 8, pp. 2080–2095, Aug. 2007. [4] B. Ahn and N. I. Cho, “Block-matching convolutional neural network forimage denoising,” CoRR abs/1704.00524, 2017.
  • 25. [5] A. Roy, J. Singha, S. S. Devi, and R. H. Laskar, “Impulse noise removal using SVM classification based fuzzy filter from gray scale images,” Signal Process., vol. 28, pp. 262–273, 2016. [6] A. Roy and R. H. Laskar, “Non-casual linear prediction based adaptive filter for removal of high density impulse noise from color images,” AEU Int. J. Electron. Commun. vol. 72, pp. 114–124, 2017. [7] G. Pok, J. C. Liu, and A. S. Nair, “Selective removal of impulse noise based on homogeneity level information,” IEEE Trans. Image Process., vol. 12, no. 1, pp. 85–92, Jan. 2003. [8] Y. Dong and S. Xu, “A New Directional weighted median filter for removal of random-valued impulse noise,” IEEE Signal Process. Lett.,vol. 14, no. 3, pp. 193–196, Mar. 2007.