Journal Publication of International Research for Engineering and Management (JOIREM)

Volume: 10 Issue: 07 | July-2024

Enhanced Analysis on Image Enrichment Methods

Ms. Pooja Uttam Waghamare
Department of Computer Science and Engineering Terna Public Charitable Trust’s College of
Engineering, Osmanabad

---------------------------------------------------------------------***---------------------------------------------------------------------
ABSTRACT— is improved by emphasizing its high frequency
content, which corresponds to edges and
The objective of Image enrichment is to process an discontinuities in the image, to enhance the fine
image so that result is more suitable than original details in the image. The underlying principle in such
image for specific application. Digital image techniques is based on the sensitivity of the human
enrichment techniques provide a lot of choices for visual system to discontinuities present in the scene.
improving the visual quality of images. Appropriate Thus, emphasizing the high frequency content would
choice of such techniques is very important. This necessarily improve the image quality. One popular
paper will provide an overview and analysis of technique that adopts such approach is the classical
different techniques commonly used for image linear unsharp masking (UM); in which contrast
enhancement. Image enhancement plays a enhancement is achieved by adding the original
fundamental role in vision applications. Recently image to a scaled highpass-filtered version of the
much work is completed in the field of images image itself. The highpass-filtered version of the
enhancement. Many techniques have previously been original image is either computed in the spatial
proposed up to now for enhancing the digital images. domain using derivative masks or in the frequency
Image enrichment is the task of applying certain domain using highpass filters [1,2].
alterations to an input image like as to obtain a more
visually pleasing image. Many images such as Several approaches have been proposed to extend
medical images, remote sensing images, electron the basic UM technique by manipulating the edge
microscopy images and even real life photographic image before it is added back to the original image.
pictures, suffer from poor contrast. Therefore it is The simplest extension is based on scaling the edge
necessary to enhance the contrast. The purpose of image by a constant value [3]. However, there are no
image enhancement methods is to increase image guidelines on how to specify the gain value.
visibility and details. Two major classifications of Additionally, this approach applies equally to all
image enhancement techniques are spatial domain regions in the image without any consideration to the
enhancement and frequency domain enhancement. level of activity. This in turn results in noise
However, these techniques bring about tonal changes amplification in smooth regions and ringing artifacts
in the images. Number of enhancement techniques near edge regions. To address the ringing artifacts
available for sketches are very few in comparison problem, the authors in [4] specify the gain scale
with the techniques available for images. The scope values that are proportional to the local activity in
of enhancement in case of composite sketches and the image regions. Basically, for a pixel in the edge
forensic sketches is left out. Therefore, it would have image, the scale value is assumed to be inversely
been better if an in-depth research work is proportional to the local standard deviation (LSD) in
undertaken. a small neighborhood around that pixel. In this
approach, ringing artifacts are significantly reduced
KEY WORDS- Image processing; Image enrichment; since edge pixels are characterized with relatively
Unsharp Masking; Nonlinear filtering high LSD values. However, amplification of noise in
smooth regions is significantly increased since
smooth regions are of low LSD values.
I. INTRODUCTION: Alternatively, the technique proposed in [5] attempts
to tackle noise amplification and edge ringing
Image contrast enhancement plays a vital role in artifacts by using a nonlinear contrast gain function
almost any image-based application. Numerous that is defined by deriving a transformation function
contrast enhancement techniques have been that maps the LSD histogram into a desired
proposed in the literature [1,2]. In one category of histogram obtained through an extension Hunt’s
these techniques, the visual quality of a digital image image model. It was proven that such gain function
 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 07 | July-2024

produces better results in terms of enhanced image Literature Survey

quality and reduced ringing and noise amplification.
Nonetheless, defining the proper gain function for an Pietro P., and Malik J. [3] proposed a new scale-space
image involves an iterative procedure that requires and edge detection algorithm using anisotropic
the user to specify three different parameters diffusion method. In their technique, the diffusion
interactively. coefficient is used in such a way to support the intra-
Images that come from a variety of microscope region smoothing as comparative to inter-region
technologies provide a wealth of information. The smoothing, so as to extract the global information
limited capacity of optical instruments combines after removing the noise from the image. Bilateral
with the noise inherent in optical imaging to make filter is introduced by Tomasi, C., and Manduchi, R.[4]
image enhancement desirable in many applications. for gray and color images. Bilateral filter is non-
Image enhancement is the process of altering the linear and non-iterative filter that preserve the edges
appearance of an image, or a subset of an image, for by mean of combining the nearby pixels values in
improved contrast or visualization of some features image. They used the technique to combine the gray
and to facilitate more accurate subsequent image levels or color based on their geometric closeness
analysis, such as classification. With image and their photometric similarity in both range and
enhancement, the visibility of selected features in an domain .F. Durand and J. Dorsey[6] presented a
image can be improved, but the inherent information technique for the display of high- dynamic-range
content cannot be increased. The design of a good images, which reduces the contrast while preserving
image enhancement algorithm should consider both detail. It is based on a two-scale decomposition of the
the specific features of interest in the image and the image into a base layer, encoding large-scale
imaging process itself. In microscopic imaging, the variations, and a detail layer. Only the base layer has
images are often acquired at different focal planes, its contrast reduced, thereby preserving detail.
different time intervals, and in different spectral Xu, L., Lu, C.et al. [8] have proposed the technique of
channels. The design of an enhancement algorithm image smoothing with L0 gradient minimization
should therefore take full advantage of this method. This method is based on the spatial changes
multidimensional information. in which a restriction is placed upon the total
number of non-zero gradients between pixels so that
A variety of image enhancement algorithms have to globally enhancing the prominent edges, even if
previously been developed and utilized for the boundaries of objects are much contract. An
microscopy applications. These algorithms can be unsharp masking algorithm is presented by
classified into two categories: spatial domain and GuandDeng[9] by using exploratory data model as a
transform domain methods. The spatial domain unified framework. The proposed algorithm is
methods include operations carried out on a whole designed to address three issues: 1) simultaneously
image or on a local region selected on the basis of enhancing contrast and sharpness by means of
image statistics. Techniques that belong to this individual treatment of the model component and
category include histogram equalization, image the residual, 2) reducing the halo effect by means of
averaging, sharpening of important features such as an edge-preserving filter, and 3) solving the out-of-
edges or contours, and nonlinear filtering. The range problem by means of log-ratio and tangent
transform domain enhancement methods manipulate operations. M. Son,Y. Lee,et al.[12] presented a novel
image information in a transform domain, such as method for enhancing details in a digital photograph,
Fourier and wavelet transform domains. In many inspired by the principle of art photography. Their
cases, interesting image information that cannot be technique provides a flexible tone transform model
separated out in the spatial domain can be isolated in that consists of two operators: shifting and scaling.
the transform domain. For example, one can often This model permits shifting of the tonal range in each
amplify certain Fourier coefficients and then return image region to enable significant detail boosting
the image to the spatial domain to highlight regardless of the original tone.
interesting image content. The wavelet transform is The total variation filter [7] uses an L1 norm based
another powerful tool that has been used for image regularization term to remove noises in images
enhancement. ,which is also considered as an edge-preserving
decomposition algorithm. Weighted Least Squares
(WLS)[13] based multi-scale decomposition
algorithm decomposes an image to two layers by
 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 07 | July-2024

solving a weighted least square optimization into groups: Histogram Equalization (HE),Tone
problem. In [11] , H. Badri, H. Yahia et al. proposed an Mapping.
accelerated iterative shrinkage algorithm to Histogram Equalization is one of the most commonly
decompose and enhance image. In [14], the L0 norm used methods for contrast enhancement. It attempts
based smoothing algorithm is introduced to a detail to alter the spatial histogram of an image to closely
enhancement scheme for fusion of differently match a uniform distribution. The main objective of
exposed images by F. Kou, Z. Li, C. Wen. In [15], it is this method is to achieve a uniform distributed
used in a visual enhancement algorithm for low histogram by using the cumulative density function
backlight displays. of the input image. The advantages of the HE include
it suffers from the problem of being poorly suited for
retaining local detail due to its global treatment of
Enrichment Methods the image small- scale details that are often
Image enrichment is the process of modifying digital associated with the small bins of the histogram are
photos so that the desired info is far better eliminated. The disadvantage is that it is not a
pertaining to show or maybe for further image suitable property in some applications such as
analysis. Image enrichment techniques can be consumer electronic products, where brightness
divided into two broad categories :- preservation is necessary to avoid annoying artifacts.
Spatial domain methods Tone Mapping is another approach of contrast
Spatial domain techniques directly deal with the enhancement techniques. In this method if we want
image pixels. The pixel values are manipulated to to output high dynamic range (HDR) image on paper
achieve desired enhancement. Spatial domain or on a display. We must somehow convert the wide
techniques like the logarithmic transforms, power intensity range in the image to the lower range
law transforms, histogram equalization are based on supported by the display. This technique used in
the direct manipulation of the pixels in the image. image processing and computer graphics to map a
Spatial techniques are particularly useful for directly set of colours to another, often approximate the
altering the gray level values of individual pixels and appearance of high dynamic range images in media
hence the overall contrast of the entire image. But with a more limited dynamic range. Tone mapping is
they usually enhance the whole image in a uniform done in the luminance channel only and in
manner which in many cases produces undesirable logarithmic scale. It is used to convert floating point
results .It is not possible to selectively enhance edges radiance map into 8-bit representation for rendering
or other required information effectively. Techniques applications. The two main aims of tone mapping
like histogram equalization are effective in many algorithm: Preserving image details and providing
images. enough absolute brightness information in low
The approaches can be classified into two categories: dynamic range tone mapped image.
Point Processing operation (Intensity transformation
function). Frequency domain systems
Spatial filter operations. Transformation or frequency domain techniques are
Point processing operations (Intensity based on the manipulation of the orthogonal
transformation function) is the simplest spatial transform of the image rather than the image itself.
domain operation as operations are performed on Frequency domain techniques are suited for
single pixel only. Pixel values of the processed image processing the image according to the frequency
depend on pixel values of original image. It can be content . The principle behind the frequency domain
given by the expression g(x,y) = T[f(x,y)] , where T is methods of image enhancement consists of
gray level transformation in point processing. They computing a 2-D discrete unitary transform of the
are especially useful for bringing out detail in Fourier image, for instance the 2-D DFT, manipulating the
transforms. transform coefficients by an operator M, and then
Direct methods define a contrast measure and try to performing the inverse transform.
improve it. Indirect methods on the other hand, The orthogonal transform of the image has two
improve the contrast through exploiting the components magnitude and phase. The magnitude
underutilized regions or the dynamic range without consists of the frequency content of the image. The
defining a specific contrast term. Contrast phase is used to restore the image back to the spatial
enhancement techniques can be broadly categorized domain. The usual orthogonal transforms are
discrete cosine transform, discrete Fourier
 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 07 | July-2024

transform, Hartley Transform etc. The transform to enhance color images by performing the
domain
main enables operation on the frequency content operations on their multiple intensity
intens components or
of the image, and therefore high frequency content on their luminance component, where image detail
such as edges and other subtle information can easily and edge information are encoded.
be enhanced.

Challenges and Future Directions

1.Image enhancement is the process of enhancing the

appearance
earance of an image, or a subset of the image, to
improve contrast or visualization of image features
of interest or to facilitate more accurate subsequent
image analysis.
2.Image enhancement can be achieved using
computational methods either in the spatiaspatial domain
or in the transform domain.
3.Spatial domain methods accomplish image
enhancement using either global operations on the
whole image, or local operations acting on a
neighborhood of each pixel. Figure 1. The developed unsharp masking
4.The operations used to increase contrast in the enhancement process
imagee include contrast stretching, clipping and
thresholding, image subtraction and averaging, and Unsharp masking enhancement process
histogram equalization and specification.
5.The operations used to sharpen image features and Unsharp masking enhancement process [1] is one of
reduce noise include spatial band-pass pass filtering, the most widespread algorithms used for image
directional and steerable
teerable filtering, and median enhancement. This technique is usually implemented
filtering. using Laplacian operator. Recently, non linear
6.If image noise is a stationary random process, operator (i.e. QV filter) has been used for image
variants of the Wiener filter can be used to reduce enhancement based on unsharp masking technique
the noise. [2]. The non linear
near operator presents an apparent
7.Nonlinear filters, such as the median filter, can problem appearing in the extraction of features from
reduce noise without blurring edges. dark areas. Thus, these areas are not well enhanced
8.Transform domain
omain methods accomplish image as brighter regions. Too, the noise is well amplified
enhancement based on computations performed in a due to the sum combination of quadratic terms for
transform domain, such as the Fourier or wavelet QV systems. The developed unsharp masking
transform domain. Often salient image features can technique, presented here, is composed by two
be more easily isolated and extracted in the branches. The first branch extracts the edge from the
transform domain than in the spatial
atial domain. input image. In the second branch we compute the
9.Commonly used Fourier domain image negative of the input image followed by the
enhancement methods include Wiener filtering, extraction of the result. The developed
dev unsharp
least-squares deconvolution, and band-pass pass filtering. masking scheme is illustrated by Figure 2.
The Wiener filter is optimal for noise removal in the
sense of minimum mean square error. In this paper the edge extraction block in Fig 1 is
10.Wavelet domain in image enhancement methods implemented using mean weighted high pass filter
leverage the advantages of multiscale image [5]. We choose the filter in order to reduce the noise
representation and nonlinear filtering. Since image amplification due to the quadratic terms in equation
edges tend to correlate spatially across multiple 2. Also, the filter extracts thinner edges from images.
scales, whereas noise does not, one can exploit this In Fig 2, the top branch of the developed approach
property and use nonlinear filtering
iltering to accentuate the edges of the negative image are extracted. In the
edge structures while suppressing noise in an image. down branch, we extract the edge from the input
11.Essentially all of the techniques developed for the image. In both branches the mean weightedwe high
enhancement of monochrome images can be applied pass filter is used for edge extraction.
 Journal Publication of International Research for Engineering and Management (JOIREM)
Volume: 10 Issue: 07 | July-2024

2. F.P.Ph. de Vries, Automatic, adaptive,

brightness indepen- dent contrast enhancement,
RESULTS Signal Processing 21 (October 1990) 169—182.
3. S. Guillon, P. Baylou, M. Najim, Robust
This paper surveys some of the areas where image Nonlinear contrast enhancement filter, Proc. 3rd
enhancement is done. This paper presents the most IEEE Internat. Conf. on Image Processing, ICIP-96,
important techniques for image enhancement in Lausanne (CH), 16—19 September 1996.
digital image processing. Although this paper did not 4. A.K. Jain, Fundamentals of Digital Image
discuss the computational cost of enhancement Processing, Prentice-Hall, Englewood Cliffs, NJ, 1989.
techniques it may play a critical role in choosing a 5. Y.H. Lee, S.Y. Park, A study of
technique for real time applications. Despite the convex/concave edges and edge-enhancing
effectiveness of each of these algorithms when operators based on the Laplacian, IEEE Trans.
applied separately, in practice one must devise a Circuits Systems 37 (7) (July 1990) 940—946.
combination of such methods to achieve more 6. S.K. Mitra, H. Li, I. Li, T. Yu, A new class of
effective image enhancement. nonlinear filters for image enhancement, Proc.
Internat. Conf. on Acoust. Speech and Signal
Processing, ICASSP-91, 1991, pp. 2525-2528.
CONCLUSION 7. A. Papoulis, Probability, Random Variables,
and Stochastic Processes, McGraw-Hill, New York,
In summary, Image enhancement, one of the most 1984.
popular ongoing topics in image processing, is 8. G. Ramponi, A Simple cubic operator for
reviewed and discussed in this survey. We first sharpening an image, Proc. IEEE Workshop on
introduced previous reviews of image enhancement Nonlinear Signal and Image Processing, 20—22 June
methods and discover that most of them are only for 1995, Neos Marmaras, Greece.
one aspect. We also figure out that the performance 9. G. Ramponi, N. Strobel, S.K. Mitra, T-H. Yu,
of existing image enhancement methods has never Nonlinear unsharp masking methods for image
been qualitatively and quantitatively evaluated in an contrast enhancement,J. Electron. Imaging 5 (3) (July
appropriate way. All these issues motivate us to 1996) 353—366.
frame this survey. Second, we introduced the existing 10. R. Gonzalez and R.E. Woods, Digital Image
classification of image enhancement methods as well Processing, Upper Saddle River, NJ, Prentice Hall,
as our proposed new classification. Then, many 2002.
commonly used image enhancement methods are 11. J.C. Russ, The Image Processing Handbook,
reviewed based on the proposed classification CRC Press, 1995.
followed by a survey of image quality metrics and 12. G. Deng, L. W. Cahill, and G. R. Tobin, “The
databases for enhanced images. After the study of logarithmic image processing model and its
comprehensive literature study, we present the application to image enhancement,” IEEE Trans. on
challenges and their potential directions of image Image Processing, vol. 4, no. 4, pp. 506–512, April
enhancement techniques. Before we draw the final 1995.
conclusion, several future perspectives for image 13. V. Digalakis, D. G. Manolakis, V. K. Ingle, and
enhancement and its evaluation are discussed. In A. K. Kok, “Automatic adaptive contrast enhancement
conclusion, universal image quality metrics and for radiological imaging,” in Proc. of IEEE Int. Symp.
databases that can be used for the performance Circuits Syst., Chicago, Illinois, May 1993.
evaluation of image enhancement approaches are 14. D. Chang and W. Wu, “Image Contrast
necessarily desired. Enhancement Based on a Histogram Transformation
of Local Standard Deviation,” IEEE Transactions on
REFERENCES: Medical Imaging, vol. 17, no. 4, pp. 518-531, Aug.
1998.
1. G. Deng, L.W. Cahill and G.R. Tobin, The study 15. A. Raji, A. Thaibaoui, E. Petit, P. Bunel, and G.
of logar- ithmic image processing model and its Mimoun, “A gray-level transformation-based
application to image enhancement, IEEE Trans. technique for image enhancement,” Pattern
Image Process. 4 (4) (April 1995) 506—512. Recognition Letters, vol. 19, no. 13, pp. 1207-1212,
Nov. 1998.