16
CHAPTER 2
LITERATURE REVIEW
2.1. REPORTED WORKS ON IMAGE ENHANCEMENT USINGEVOLVED OPERATORS
[6] Presented a circuit representation technique for automated circuit
design. The applications are mainly in the areas of classification and control
when complete circuit design is applied. There are also some examples of circuit
parameter tuning. It is based on digital gate level technology using GA as the
evolutionary algorithm. However, promising results are given for analog designs,
where evolution is used to find optimal parameters for analog components.
[14] Proposed image Filter Design with Evolvable Hardware. It
introduces a new approach to automatic design of image filters for a type of
noise. The approach employs evolvable hardware at simplified functional level
and produces circuits that outperform conventional designs. If an image is
available both with and without noise, the whole process of filter design can be
done automatically, without the influence of a designer.
[19] Proposed virtual reconfigurable Circuits for Real-World Applications
of Evolvable Hardware. The evolved image filters use Cartesian genetic
programming (CGP) applied at the functional level. Furthermore, the hardware
implementation of CGP was proposed for FPGAs. However, in his approach
image filters were evolved only by using a virtual reconfigurable circuit
simulated in software that could eventually be implemented on the top of a
conventional FPGA.
[20] Presented an accelerated image processing architecture on FPGAs
with parallel processing elements. A convolution operation is implemented in
17
FPGA to be applied for real-time image processing. It has also been proposed to
evolve image filters in reconfigurable logic.
[31] Presented a universal noise removal algorithm with an impulse
detector. An important problem of image processing is to effectively remove
noise from an image while keeping its features. There are two noise models that
can be used to represent most noise in images: additive Gaussian noise and
impulse noise. Additive Gaussian noise is characterized by adding to each image
pixel a value with a zero-mean Gaussian distribution. Such noise is usually
introduced during image acquisition.
[37] Proposed digital Filter Design using Evolvable Hardware Chip for
Image Enhancement. Images acquired through modern cameras may be
contaminated by a variety of noise sources (e.g. photon or on chip electronic
noise) and also by distortions such as shading or improper illumination.
Therefore, a pre-processing unit has to be incorporated before recognition to
improve image quality.
[40] Presented reducing the area on a chip using a bank of evolved filters.
It is applying EA in image filtering can be separated into the following two
categories: (i) parameter tuning for improving performance of the existing filters,
(ii) designing a new structure filter by EA. Both types of works have an explicit
target: an optimal filter circuit. Comparing with these works, the proposed image
filter based approach possesses the different features: noise cancellation is
performed only on the noise candidates and noise free pixels will not be
changed. Thus, more image edge detail can be preserved and computation effort
will be reduced.
[41] Presented evolvable reconfigurable hardware framework for edge
detection. Systems on Reconfigurable Chips contain rich resources of logic,
memory and processor cores on the same fabric. This platform is suitable for
18
implementation of Evolvable Reconfigurable Hardware Architectures (ERHA).
This architecture is a suitable candidate for implementation of early-processing
stage operators of image processing such as filtering and edge detection.
[49] Presented reconfigurable hardware implementations for lifting-based
DWT image processing algorithms. This scheme presents advantages over the
convolution-based approach, for instance it is very suitable for parallelization.
This paper presents two new FPGA-based parallel implementations of the DWT
lifting-based scheme, (i) uses pipelining, parallel processing and data to increase
the speed up of the algorithm, and (ii) a controller is introduced to deploy
dynamically a suitable number of clones according to the available hardware
resources on a targeted environment.
[59] Proposed EHW Architecture for Design of Adaptive Median Filter
for Noise Reduction. A new technique for the design of Adaptive Median Filter
within an Evolvable hardware framework, using genetic algorithm (GA), aimed
at removing the impulse noise from the image and reducing distortion in the
image is presented. It reduces the number of generations required to provide time
bound optimal filter configuration and to improve the quality of the filter
designed.
[62] Proposed application of partial reconfiguration of FPGAs in image
processing. FPGA based hardware accelerators have been more and more widely
used in different kind of applications. As compared to other solutions and the
direct hardware implementation, the advantage of the FPGA devices is their
flexibility that arises from their programmable nature. In addition to this, some
FPGA devices also support partial dynamic reconfiguration.
[86] Presented an efficient Image Noise Removal and Enhancement
Method. It presents a new method to remove noise and enhance the image with
the help of partial unsharp masking and conservative smoothing. In this method,
19
unsharp masking is applied in partial way for detection of the edges and
boundary lines in the image and then a conservative smoothing operation is
applied on the selected areas to remove undesirable edges which represents the
salt and pepper noise.
[90] Proposed image enhancement based on Improved Genetic Algorithm
and Lifting Wavelet Method. This algorithm improves crossover operation
algorithm and utilizes average displacement method for mutation operation.
Probabilities of crossover and mutation are selected adaptively. It decided the
fitness function and has implemented multi-thread design. The algorithm
optimizes the prediction and updating operator of lifting wavelet by means of
genetic algorithm.
[99]Presented architecture for binary mathematical morphology
reconfigurable by genetic programming. The mathematical morphology supplies
powerful tools for low level image analysis, with applications in robotic vision,
visual inspection, medicine, texture analysis and many other areas. Many of the
mentioned applications require dedicated hardware for real time execution. The
development of a novel reconfigurable hardware using logical and
morphological instructions generated automatically by a linear approach based
on genetic programming is proposed.
[100] Presented reconfigurable hardware objects for image processing on
FPGAs. Embedded systems require high level of abstraction is required during
the development process. High abstraction methods simplify implementation of
complex computation systems and shorten the time to market. It represents an
implementation of a graphic computing element (GCE) which can be used as a
runtime parameterized building block in image processing applications in
FPGAs.
20
[92] Proposed an edge Enhancement Algorithm using Wavelet Transform
for Automatic Edge Detection in Synthetic aperture radar (SAR) Images. It
presents a novel technique for automatic edge enhancement and detection in
SAR images. The characteristics of SAR images justify the importance of an
edge enhancement step prior to edge detection. Therefore, it presents a robust
and unsupervised edge enhancement algorithm based on a combination of
wavelet coefficients at different scales.
[94] Proposed image enhancement and denoising based on structure
self-similarity and wavelet transform coefficients. Image denoise is a very
important method in image quality improvement and the image quality
evaluation is another important aspect in image processing. In this work, a novel
enhancement algorithm based on wavelet transform (WT) and structure self-
similarity (SSS) is introduced, which was attested to algorithm effective for
enhancement and confining random noise.
[110] Proposed an efficient run-time task allocation in reconfigurable
multiprocessor system-on-chip with network-on-chip. Due to the advancement
of VLSI (Very Large Scale Integrated Circuits) technologies, it can put more
cores on a chip, resulting in the emergence of a multicore embedded system. It
also brings great challenges to the traditional parallel processing in improving
the performance of the system with increased number of cores.
2.2. REPORTED WORKS ON SURFACE ROUGHNESS
[32] Presented a problem of emphasizing the features of surface
roughness by Discrete Wavelet Transform. The detection of the roughness
features by means of the 3D reconstruction, based on photometric stereo
techniques, an important problem is the elimination of the brightness variation
due to different light conditions which can alter the response. The level of
brightness depends on many factors as well as the homogeneity of reflection
21
properties of the material or its physical continuity and the surface smoothness or
roughness.
[29] Presented an application of digital image magnification for surface
roughness evaluation using machine vision. A machine vision system has been
utilized to capture the images and then the quantification of the surface
roughness of machined surfaces (ground, milled and shaped) is done by the
application of regression analysis. Subsequently, original images have been
magnified using Cubic Convolution interpolation technique and improved (edge
enhancement) through Linear Edge Crispening algorithm.
[63] Presented restoration of blurred images for surface roughness
evaluation using machine vision. The surface roughness of uniformly moving
machined surface (grinding, milling) using machine vision technique is
evaluated. In the case of moving surfaces the images are likely to blur due to the
relative motion between the CCD camera and the object to be captured. Hence,
the degraded image has to be restored by removing distortion due to motion
before subsequent analysis.
In [2], four methods that yield mathematical measures to analyze the
precision of surfaces of manufactured parts is investigated. In terms of precision
manufacturing, measures provide the potential of detecting and improving
surface errors in high-precision product geometry. The average energy is given
by the eigenvalues of the covariance matrix. The covariance matrix contains the
statistical properties of the original data set. In [2], they investigate the feasibility
of finding characteristic measures of precision for precision-ground surfaces.
[3] Includes intuitive properties like roughness, granulation and
regularity. To obtain features which reflect scale-dependent properties, one can
extract a feature from each sub image separately. The transforms retain
localization in both space and frequency, which makes it easy to compute
22
multiscale features locally. Thus rotation invariant features, which preferably
still reflect the anisotropy, need to be constructed. Colour images are typically
represented by RGB tristimulus values which correspond to three colour bands.
A straight forward way to process colour textures is by performing a gray level
decomposition on every component image. They have given an overview of the
application of wavelet multi resolution image analysis to texture.
In [12] a laser based system was used to scan the surfaces of 6 steel
sheets. The resulting waveforms were pre-processed and then they were
represented by a set of feature vectors. The light which is reflected contains
information regarding the surface profile of the surface. Then the experimental
procedures will be outlined and the results obtained will be discussed. [12] is due
to the fact that Ra is a measure of amplitude while the variance is a measure of
signal power. In [12], study a commercial optical profilometer was used to scan
the surfaces of steel sheets with 6 different known average surface roughness
(Ra) values.
Standard roughness measurement procedures depend heavily on stylus
instruments, which have only limited flexibility in handling different parts. [64]
is organized as follows. The definition of surface roughness in manufacturing
fields is first described more formally. There are various simple surface
roughness amplitude parameters such as roughness average, root mean square
roughness, and maximum peak to valley roughness, etc. Various spindle speeds,
feed rates, and depths of cut were tested. During the machining, an accelerometer
sensor was used to measure the vibrations. The aim of modeling the end milling
process is predicting the surface roughness of a work-piece machined.
Conventional stylus techniques, though powerful, have their own
limitations. These include the resolution of the stylus, and the damage caused by
the moving diamond stylus when tracing profiles on soft materials. The
diffraction pattern for a rectangular aperture, which is normally illuminated with
23
a plane wave, is concentrated principal lying two directions coinciding with the
sides of the aperture, and in each of these directions it corresponds to the width
of the aperture in that direction [1]. From [1], it was observed that the optical
diffraction technique gave good results for turned components of medium
roughness, in spite of the limitations cited above. The use of a helium-neon laser
beam of smaller diameter and a very smooth knife edge could improve the
results further.
With hard turning, which is an attractive alternative to existing grinding
processes, surface quality is of great importance. Signal processing techniques
were used to relate work piece surface topography to the dynamic behaviour of
the machine tool [45]. Hard turning is an established industry process in industry
for finish machining of a wide range of hardened steel work pieces. Hard turning
allows manufacturers to simplify their processes and still achieve the desired
surface finish quality. From [45] they observed that the work piece topography
generated by hard turning is affected by the feed rate, macro tool geometry,
micro tool geometry (tool wear), and machine tool vibrations, etc. The frequency
components of the profile correspond to these factors.
[15] Reports that the deeper a valley, the darker the corresponding pixel,
the higher a peak, the brighter the corresponding area in the image. This
approach is commonly used to describe the continuous wavelet transform (CWT)
for which the mother wavelet can be explicitly expressed. Hence, the wavelet
toolbox allows the decomposition of surfaces into form, waviness and roughness
components well appreciated by mechanical engineers. Those features can also
be quantified according to both the shape of the corresponding peak and its
height. The standard wavelet transform allows separating the different frequency
components of an image. Indeed, characterization by FNWT seems to be a
promising strategy in the field of surface roughness characterization.
24
[46] Concluded that CWT can be useful for the analysis of the roughness
products generated by cutting and abrasive machining processes. This situation
leads to the conclusion that other tools must be used for proper analysis of non
stationary profiles. The sample length was set to 0.8 mm and measurement
length to 4.8 mm. Mean values and confidence intervals of vertical and hybrid
parameters values are decreasing with every step of technological process, but
the horizontal parameters change their values in a different way. The results
reported in [46] are also confirmed by CWT matrix with the use of wavelet
“Mexican hat”. CWT using basic Morlet wavelet allows evaluation of the length
of profile constituent wavelets but information about their amplitudes is not
precise.
The work in [7] is based upon wavelets theory, a novel reference for
evaluating surface roughness is proposed here, wherein the surface roughness
can be separated from the actual surface profile f(t). With the rapid development
of high technology, the quality requirement of many manufactured surfaces are
getting more and more strict in the fields of machine building, electronics,
optics, and biomedical engineering, even in domestic industry; therefore, how to
evaluate the surface feature reasonably is becoming increasingly important.
From [7] they observed that the wavelet references are natural and smooth lines
or surfaces without algebraic expression. It is unique. The evaluation precision
of surface roughness by wavelet reference could be higher than that by the
classic reference lines. They are smooth arithmetical mean lines of the profile.
In [8], the paper proposes a new strategy for surface roughness analysis
and characterization based on wavelets. A three-step algorithmic proposed to
perform a task of surface roughness discrimination between surface texture
images coming from eight different engineering processes. Real samples coming
from British Standards roughness comparison specimens were measured in 2D
using an optical measurement system .All the measured surfaces had been
manufactured to have the same arithmetic mean deviation of the surface
25
roughness value Ra 0:8lm. In [8] it can be seen that the three texture scanning
methods offer good clustering efficiency especially when using a clustering
method based on the cluster analysis. To be more precise, it can be seen that
scanning by CWT and standard DWT gives better performances. This comes
from the fact that the scaled DWT scanning is not orientation selective and hence
less efficient for texture analysis where texture orientation is a point of
importance.
2.3 REPORTED WORKS ON IMAGE DENOISING AND WAVELET
TRANSFORM
[4] Presented wavelet based image denoising using a Markov Random
Field a priori model. Hidden Markov Models (HMM) models are efficient in
capturing inter-scale dependencies, whereas Random Markov Field models are
more efficient to capture intrascale correlations. The complexity of local
structures is not well described by Random Markov Gaussian densities whereas
Hidden Markov Models can be used to capture higher order statistics.
[11] Proposed that the multiscale Canny-Deriche operator gives the best
performance of all models and they evaluate the performance along with pre-
processing techniques using wavelet transform applied to face image as follows:
The linear auto-associate model applied to face images is presented. Since,
autoassociators are generally interpreted as content addressable memories; their
performance is evaluated by comparing the output of the system with a test
pattern which can be a copy or a degraded version of one of the patterns
previously learned by the system.
[30] Presented salt-and pepper noise removal by median-type noise
detectors and detail preserving regularization. It was reported that
malfunctioning pixels in camera sensors, faulty memory locations in hardware or
transmission of the image in a noisy channel are some of the common causes for
impulse noise.
26
[39] Presented thresholded Weighted Median Filters for Ringing
Reduction in Processed Images. It presents how thresholded weighted median
filters (WMFs) can significantly improve visual as well as objective quality of
images affected by ringing. Ringing is identified by its structural properties and
WMFs are chosen according to these structures. Since WMFs are concerned with
the relative values of their input but do not consider its absolute values, a limited
at the filter output is explicitly required.
[50] Proposed application of two-dimension wavelet transform in image
process of pets in stored grain. The wavelet transform is the localization analysis
of time and frequency and it can multi-scale refine the signal by calculating of
flex and transition. It presents a method of using the wavelet transform to detect
the image of pests in stored grain edge based on the multi-scale analysis of the
wavelet transform in the image processing field. The method acquires the
information of the image edge by detecting the image local maxima of the two-
dimension wavelet transform.
[53] Proposed de-noising of natural images corrupted by Gaussian noise
using wavelet techniques and is very effective because of its ability to capture
the energy of a signal in few energy transform values. Investigates the suitability
of different wavelet bases and the size of different neighbourhood on the
performance of image de-noising algorithms in terms of PSNR. The next level of
wavelet transform is applied to the low frequency sub band image LL only. It is
a shrink or kill rule. However, there was a slight improvement in the PSNR of
the reconstructed image using wiener filtering. The de-noised image was
sometimes unacceptably blurred and lost some details. The different wavelet
bases are used in all methods. It was found that fourth level decomposition gave
optimum results.
[52] Presented details preserving median based filter for impulse noise
removal in digital images. A new nonlinear filter called detail preserving median
27
based filter for removing salt and pepper noise and random valued impulse noise
with edge and detail preservation is presented. The proposed method first detects
the impulse pixel based on threshold values and then the corrupted pixels are
replaced by the median value of the uncorrupted pixels in the filtering window.
[54] Presented simple adaptive median filter for the removal of impulse
noise from highly corrupted images. It presents a simple, yet efficient way to
remove impulse noise from digital images. This novel method comprises to
detect the impulse noise in the image. It is based on the intensity values, the
pixels are roughly divided into two classes, which are “noise-free pixel” and
“noise pixel”. Then, stage is to eliminate the impulse noise from the image.
[55] Presented a new kind of weighted median filtering algorithm used
for image Processing. The new algorithm first determines noise points in image
through noise detection, then adjusts the size of filtering window adaptively
according to number of noise points in window, the pixel points in the filtering
window are grouped adaptively by certain rules and gives corresponding weight
to each group of pixel points according to similarity, finally the noise detected
are filtering-treated by means of weighted median filtering algorithm.
[58] Proposed an Improved Switching Median Filter for Impulse Noise
Removal. An Improved progressive switching median filter proposed for salt-
and-pepper impulse noise removal from digital images. Results of comparative
analysis of this algorithm with other filters for impulse noise removal show a
high efficiency of this approach relatively to other ones.
[57] Proposed the effect of Input Limiting on Linear and Nonlinear
Filters for the Removal of Impulsive Noise. It investigates the effects of input
limiting on the performance of linear and nonlinear filters when the input signal
is contaminated by impulsive noise modeled as an alphastable random process.
The nonlinear filters chosen are from the median family filters: median and
28
recursive median filter and Wiener filter from the linear filter group. The
hypothesis being tested in this paper is that front-end limiting will increased the
performance of median filter when the noise was extremely impulsive.
[61] Presented an efficient non-linear cascade filtering algorithm for
Removal of High Density Salt and Pepper Noise in Image and Video sequence.
This method consists of two stages to enhance the filtering. (i) Decision based
Median Filter (DMF) which is used to identify pixels likely to be contaminated
by salt and pepper noise and replaces them by the median value.
(ii) Unsymmetric Trimmed Filter, either Mean Filter (UTMF) or Midpoint Filter
(UTMP) which is used to trim the noisy pixels in an unsymmetrical manner and
processes with the remaining pixels.
[74] Presented video denoising Using Motion Compensated 3-D Wavelet
Transform with Integrated Recursive Temporal Filtering. The motion-
compensated temporal wavelet transform is first performed on a sliding window
of video frames consisting of previously denoised frames and the current noisy
frame. The 2-D spatial wavelet transform is then performed on the temporal sub
band frames, thus, realizing a 3-D wavelet transform.
[75] Proposed image de-noising algorithm study and realization based on
wavelet analysis. It collects blur image with missing information from an
imaging system. The algorithm based on wavelet analyses does not need any
transcendental information of the image or the image size to estimate the de-
nosing limits and even does not need the information of square difference .It has
a function to reduce image noise blindly.
[85] Presented an effective adaptive median filter algorithm for removing
salt & pepper noise in images. It solved the problem of the simplified Pulse
Coupled Neural Network model in image filtering. The simplified model is
proved to fail to detect pepper noise using the method of reduction ad absurdum
29
and the model is improved using the method of divide and rule. Finally, the
adaptive median filter algorithm is achieved by detecting the pollution level of
the image.
[73] Presented application of modified adaptive median filter for impulse
noise. In this paper, based on the statistical features of the image, a modified
adaptive median filter (MAMF) for removal of impulse noise, especially for the
high-density impulse noises is proposed. To avoid the adaptive median filter
(AMF) and the adaptive threshold median filter (ATMF), this method has been
designed by combining the AMF with the Decision-Based Algorithm (DBA).
[69] Presented the mathematical Programming Problem of Total
Variation Image Denoising Model. Image denoising is an image processing
problem, which has a wide use. Total Variation image denoising model is one of
the best models at the present time. According to its features, it proposes three
mathematical programming models with guaranteed global optimum.
[71] Presented the Image Edge Detection Algorithm based on Wavelet
Denoising and Mathematics Morphology. The improved wavelet semisoft
threshold method is used to suppress noise of image, where the algorithm of
Bayes threshold is adopted to calculate the value of threshold. A multi-scale and
multi-structural elements morphological edge detection algorithm with entropy
weights is presented and compared with two more popular morphological edge
detection algorithms from three aspects (run times, SNR and precision).
[76] Presented an improved threshold denoising algorithm based on inter-
scale dependency of wavelet. An efficient method based on threshold denoising
algorithm to remove the noise in the image. To the disadvantages of the unified
threshold denoising method, which causes the image fuzzy distortion because of
“over-killed”, by using inter-scale dependency of wavelets coefficients, some
30
edge information that “overkilled” by the unified threshold are extracted and
reserved.
[78] Presented a new Adaptive Switching Median (ASWM) filter for
removing impulse noise from corrupted images. The originality of ASWM is that
no apriori Threshold is needed as in the case of a classical Switching Median
filter. Instead, Threshold is computed locally from image pixels intensity values
in a sliding window. It provides better performance in terms of PSNR and MAE
than many other median filter variants for random-valued impulse noise.
[79] Presented a median Filter Method for Image Noise Variance
Estimation. Image noise estimation is of crucial importance for the computer
vision algorithm, for the algorithm parameter is always adjusted to account for
the variations in noise level over the captured images. A median filter method is
provided for the image noise variance estimation in the paper. The image was
processed with a group of high pass digital filters constructed by several finite
difference operators with different orders.
[80] Proposed a superior methodology based on improved tolerance
based selective arithmetic mean filtering technique for the detection and removal
of Salt and Pepper Noise on corrupted images is presented in this paper. The
function of the proposed filtering technique is to detect and remove the noisy
pixels and restore the noise free information.
[81] Proposed a new and efficient algorithm for the removal of high
density salt and paper noise and videos. The existing non-linear filter like
Standard Median Filter (SMF), Adaptive Median Filter (AMF), Decision Based
Algorithm (DBA) and Robust Estimation Algorithm (REA) shows better results
at low and medium noise densities. At high noise densities, their performance is
poor. The new algorithm has lower computation time when compared to other
standard algorithms.
31
[82] Proposed removal of salt-and-pepper noise based on compressed
sensing. The key point of the method is the sparsity used to reconstruct the
whole image based on partial noise-free pixels. It demonstrates the better
performance of the proposed method compared to the existing modified median-
type filters.
[88] proposed wavelet denoising Double-Threshold Optimization Method
and Its Application. A method based on ant colony algorithm is given for
optimizing wavelet de-noising double-threshold. The optimization interval and
the objective function are chosen according to the difference of autocorrelation
coefficient, which belong to signal’s wavelet coefficient and noise’s wavelet
coefficient respectively. The optimal upper threshold and lower threshold are
calculated by ant colony algorithm.
[98] Presented real-time dynamically reconfigurable 2-D filter banks. It
is based on separable one-dimensional filters. At the lowest level, each 2D filter
is implemented using dynamic reconfiguration between two one-dimensional
filters. Then, at a higher level, filter banks are implemented using dynamic
partial reconfiguration of efficient 1D filter blocks (based on distributed
arithmetic).
[101] Proposed reconfigurable hardware for median filtering for image
processing applications. Median filter is a non-linear filter used in image
processing for impulse noise removal during morphological operations, image
enhancement and other image processing operations. It finds its typical
application in the situations where edges are to be preserved for higher level
operations like segmentation, object recognition etc. Real-time applications, such
as video and high speed acquisition cameras often require fast algorithms for
processing.
32
[91] Proposed the framework for reconfigurable architecture based
floating point Discrete-Wavelet Transform computation algorithm. The discrete
wavelet transform has taken its place at the forefront of research for the
development of signal and image processing applications. Hence, this work is
proposed on the design of hardware for the computation of Floating Point
Discrete Wavelet Transform using Harr wavelet. The hardware was implemented
using FPGA with gate level.
[93] Proposed image denoising using multi-scale thresholds method in the
wavelet domain. Images often contain noise due to the capturing devices
environment. For further processing, compression, fractal, etc the image
denoising is necessary. Wavelet analysis plays a very important role in image
denoising. It improves the wavelet thresholding method by using multi-scale
thresholds and a new thresholding function and in case of large noise, a median
filter is suggested.
[95] Proposed a modified Retinex Algorithm based on wavelet
transformation. Retinex method mainly consists of two steps: estimation and
normalization of illumination. The illumination is estimated as a smooth version
of input image using low-pass filters. Some high-frequency components of
image will inevitably be lost in the filtering processing, and images will lose
details and information, correspondingly.
[97] Presented switching bilateral filter with a texture/noise detector for
universal noise removal. In this work, for detection, a sorted quadrant median
vector (SQMV) scheme, which includes important features such as edge or
texture information is proposed. This information is utilized to allocate a
reference median from SQMV, which is in turn compared with a current pixel to
classify it as impulse noise, Gaussian noise, or noise-free.
33
[108] proposed a new Adaptive Weight Algorithm for Salt and Pepper
Noise Removal that consists of two major steps, (i) to detect noise pixels
according to the correlations between image pixels, (ii) use adaptive methods
based on the various noise levels. For the low noise level, neighbourhood signal
pixels mean method is adopted to remove the noise and for the high noise level,
an adaptive weight algorithm is used.
[106] proposed a comprehensive Analysis and Parallelization of an
Image Retrieval Algorithm. The advent of multi-core hardware has opened new
opportunities to improve the effectiveness of multimedia data processing. It
make a comprehensive analysis on different potential parallelism, including
pipeline parallelism, task parallelism at both scale level and block level, data
parallelism, and their combinations, in a typical image retrieval algorithm called
SURF, which is the core algorithm of many multimedia (i.e., image and video)
retrieval applications.
[27] Describes the new approach to construct the best tree on the basis of
Shannon entropy. The proposed algorithm provides a good compression
performance. Basis functions are obtained from a single photo type wavelet
called the mother wavelet by dilation (scaling) and translation (shifts). These sets
are divided into four parts such as approximation, horizontal details, vertical
details and diagonal details. They have implemented the proposed algorithm,
wavelet packet best tree using Shannon entropy. In [104], the Wavelet Packet
Best Tree using Shannon entropy has been presented. An extensive result has
been taken on different images.
34
2.4 REPORTED WORKS ON CURVELETS, CONTOURLETS AND
2D-PCA
[34] Proposed texture orientation and anisotropy calculation by Fourier
transform and Principal Component Analysis. It proposes a simple method to
calculate a texture angle of orientation and degree of anisotropy. The principle of
the algorithm is to calculate the image Fourier transform modulus and then to
characterize the distribution of this spectrum around the zero frequencies by
Principal Component Analysis. The algorithm returns both the angle of
orientation and an index of confidence.
[84] Presented image denoising using contourlet and two-dimensional
principal component analysis (2DPCA). The noise image was decomposed using
Contourlet by forming directional sub bands. The 2DPCA is then carried out to
estimate the threshold for the image blocks in high frequency sub bands. Thus
the soft thresholding shrinkage can be employed on the Contourlet coefficients
without estimating the noise variance. The denoising algorithm is validated by
numerical study on two images.
[70] Presented a new method based on Curvelets Transform for Image
Denoising. Curvelet transform that combines both Window Shrink and Bayes
Shrink were reported. Though the Wavelet transform can perform job well, it has
a low Resolving rate in high frequency area and it also lacks of the direction in
dealing with images. Curvelet transform have an efficient way of representing
the line and surface property of image.
[96] proposed wavelets, Curvelets and Wave Atoms for Image Denoising.
The images usually bring different kinds of noise in the process of receiving,
coding and transmission. The wavelet transform, Curvelet transform and wave
atom were used for denoising of a image with Gaussian noise. The digital
implementations of three newly developed multi-scale representation systems
35
were proposed. The Curvelet transform and wave atom frame are two kinds of
new multi-scale transform after that is based on wavelet transform.
[109] Presented performance evaluation of Curvelet and Wavelet based
Denoising Methods on Brain Computed Tomography Images. It presents the
evaluation of the effect of noise reduction techniques on the brain Computed
Tomography (CT) images. In particular, multiscale geometric denoising methods
based on Curvelet transform are used and compared with wavelet based
methods. It shows that cycle spinning based Curvelet transform method
outperforms the wavelet based methods not only for the suppression of noise but
also for preservation of fine details, edges and allow the use of a low dose brain
CT images.
[67] Presented image enhancement by fusion in contourlet transform. The
image enhancement algorithms work on a single image. Their performance is
limited to the capacity of the sensor by which the image is taken. It provides the
necessary enhancements to composite image approach for enhancing still
images. The approach proposed combines the relevant features of the input
images and produce a composite image which is rich in information content for
human eye.
[65] Presented gray and color image contrast enhancement by the
Curvelet transform. It has one way to solve the problems is to apply image
enhancement original single image. A lot of algorithms developed in this area
and their performance is limited with the performance of the sensors in which the
image is taken. Either due to design or observational constraints a single image
approach usually fails in providing the necessary enhancements.
[66] Proposed multi-sensor image enhancement and fusion for vision
clarity using contourlet transform. It composites image approaches employ pixel
fusion methods that has advantage of pixel fusion is the images used contain
36
the original information. Furthermore, the algorithms are rather easy to
implement and time efficient. An important pre-processing step in pixel based
fusion methods is image registration, which ensures that the data at each source
is referring to the same physical structures.
[16] presented although the wavelet transform has been proven to be
powerful in many signal and image processing applications such as compression,
noise removal, image edge enhancement and feature extraction; wavelets are not
optimal in capturing the two-dimensional singularities found in images.
Therefore, several transforms have been proposed for image signals that have
incorporated directionality and multi-resolution and hence, more efficiently
capture edges in natural images.
[17] Presented directional multiscale modeling of images using the
contourlet transform. In the case of the contourlet transform can assume two
different parent child relationships depending on the number of directional
decompositions in the contourlet subbands. If the two successive scales in which
the parent and children lie have the same number of directional decompositions
then the parent and children would lie in the corresponding directional subbands.
2.5 INFERENCE FROM LITERATURE REVIEW
From the literature review surveyed, it is observed that the functional
behaviour of manufactured surfaces is influenced by errors such as roughness,
waviness and form errors that are present on the surface and these errors
influence the functional behaviour. Also, it has been observed that current
evolutionary techniques based filtering schemes has practical limitation when
applied for complex real world problems. The search spaces can become vast for
large circuits and a greater deal of research needs to be directed at scalability. In
this work, it is presented that, one can still evolve circuits with limited
interactions that can be used by traditional designers as building blocks for larger
circuits. Initial research involved evolving circuits at a very high primitive gate
37
level and results obtained using this approach showed that evolved circuits were
less useful for more demanding commercial applications. To overcome this
problem a function-level evolution is proposed in this work and domain
knowledge is used to select high level computational units, which can be
represented directly in the chromosome. Previous reported works on machined
image enhancement depends on model based approach as compared to the EHW
based image enhancement filter using coordinate logic operators and functional
level evolution concept presented in this work. Image enhancement schemes
reported so far are dependent on the noise frequency band and machining
specifications. On the contrary, in this work the presented evolutionary operator
and 2D transforms based schemes have the advantage, that it is independent of
the frequency band in which the noise affects the image and specifications of
milling and grinding. In most of the analogous works surveyed and presented,
the use of wavelet transform for designing filter to extract the image features as
well to denoise the image is less suited for image alignment. However, in this
work, it is made suited to detect a highly anisotropic element that includes image
alignments by proposing a 2D-PCA based enhancement scheme. As a result, the
presented feature extraction technique can be adapted generically in machine
vision applications.
Top Related