1. Describe the fundamental steps in image processing Multiresolution Processing : Wavelets are the foundation 2.

n Processing : Wavelets are the foundation 2. Explain the smoothing and sharpening filters in the . Edges are better preserved in this filtering. Types of Order
using suitable block diagram. Image Processing is for representing images in various degrees of resolution. context of spatial filtering. Use suitable examples. statistics filter: (1) Minimum filter: 0th percentile filter
a method to convert an image into digital form and perform Images subdivision successively into smaller regions for Spatial Filtering technique is used directly on pixels of an is the minimum filter. The value of the center is replaced
some operations on it, in order to get an enhanced image or data compression and for pyramidal representation. 6. image. Mask is usually considered to be added in size so by the smallest value in the window. (2) Maximum filter:
to extract some useful information from it. Block Diagram Compression : Compression deals with techniques for that it has specific center pixel. This mask is moved on the 100th percentile filter is the maximum filter. The value of
of Digital Image Processing System : 1. Image reducing the storage required to save an image or the image such that the center of the mask traverses all image the center is replaced by the largest value in the window.
Acquisition : This is the first step or process of the bandwidth to transmit it. Particularly in the uses of the pixels. There are two types of spital filters : a) Smoothing (3) Median filter: Each pixel in the image is considered.
fundamental steps of digital image processing. Image internet, it is very much necessary to compress data. 7. Spatial Filter and b) Shapening Spatial Filter a) First neighboring pixels are sorted and original values of
acquisition could be as simple as being given an image that Morphological Processing : Morphological processing Smoothing Spatial Filter: Smoothing filter is used for the pixel is replaced by the median of the list. (b)
is already in digital form. Generally, the image acquisition deals with tools for extracting image components that are blurring and noise reduction in the image. Blurring is pre- Sharpening Spatial Filter: It is also known as derivative
stage involves per-processing, such as scaling, etc. 2. useful in the representation and description of shape. 8. processing steps for removal of small details and Noise filter. The purpose of the sharpening spatial filter is just the
Image Enhancement : Image enhancement is among the Segmentation : Segmentation procedures partition an Reduction is accomplished by blurring. Types of opposite of the smoothing spatial filter. Its main focus in
simplest and most appealing areas of digital image image into its constituent parts or objects. In general, Smoothing Spatial Filter: i. Linear Filter (Mean Filter) on the removal of blurring and highlight the edges. It is
processing. Basically, the idea behind enhancement autonomous segmentation is one of the most difficult tasks ii. Order Statistics (Non-linear) filter These are explained based on the first and second order derivative. First order
techniques is to bring out detail that is obscured, or simply in digital image processing. A rugged segmentation as following below. i ) Mean Filter: Linear spatial filter derivative: Must be zero in flat segments. Must be non
to highlight certain features of interest in an image. Such as procedure brings the process a long way toward a is simply the average of the pixels contained in the zero at the onset of a grey level step. Must be non zero
changing brightness & contrast, etc. 3. Image Restoration successful solution of imaging problems that require neighborhood of the filter mask. The idea is replacing the along ramps. First order derivative in 1-D is given by: f'
: Image restoration is an area that also deals with improving objects to be identified individually. 9. Representation value of every pixel in an image by the average of the grey = f(x+1) - f(x). Second order derivative: Must be zero
the appearance of an image. However, unlike and Description : Representation and description almost levels in the neighborhood define by the filter mask. Types in flat areas. Must be zero at the onset and end of a ramp.
enhancement, which is subjective, image restoration is always follow the output of a segmentation stage, which of Mean filter: (1) Averaging filter: It is used in reduction Must be zero along ramps. Second order derivative in 1-
objective, in the sense that restoration techniques tend to be usually is raw pixel data, constituting either the boundary of the detail in image. All coefficients are equal. (2) D is given by: f'' = f(x+1) + f(x-1) - 2f(x) .  Example
based on mathematical or probabilistic models of image of a region or all the points in the region itself. Choosing a Weighted averaging filter: In this, pixels are multiplied by of Median Filter : It replaces the value at the center by
degradation. 4. Color Image Processing : Color image representation is only part of the solution for transforming different coefficients. Center pixel is multiplied by a higher The median pixel value in the neighborhood, (i.e. the
processing is an area that has been gaining its importance raw data into a form suitable for subsequent computer value than average filter. ii) Order Statistics Filter: It is middle element after they are sorted) Median filters are
because of the significant increase in the use of digital processing. Description deals with extracting attributes that based on the ordering the pixels contained in the image area particularly useful in removing impulse noise (also known
images over the Internet. This may include color modeling result in some quantitative information of interest or are encompassed by the filter. It replaces the value of the center as salt-and-pepper noise). Salt = 255, pepper = 0 gray
and processing in a digital domain etc. 5. Wavelets and basic for differentiating one class of objects from another. pixel with the value determined by the ranking result. levels. In a 3×3 neighborhood the median is the 5th largest
Multiresolution Processing : Wavelets are the foundation 10. Object recognition : Recognition is the process that Edges are better preserved in this filtering. Types of Order value, in a 5x5 neighborhood the 13th largest value, and so
for representing images in various degrees of resolution. assigns a label, such as, vehicle to an object based on its statistics filter: (1) Minimum filter: 0th percentile filter on.
Images
3. Whatsubdivision successivelybetween
is noise? Differentiate into smaller Gaussianregions for
noise ,Gaussian noise.
descriptors. 11. -Knowledge
Causes: Impulsive noise can bemay
Base : Knowledge causedbe 4. Explain
is the minimumthe chromaticity
filter. The value diagram using isa replaced
of the center labeled As seen in the figure the chromaticity diagram is formed by
and
data impulsive
compression noise?and for pyramidal Noiserepresentation.
in the context of 6. by simple
as errorsas during
detailing image
regions transmission,
of an image where hardware the diagram. Reason
by the smallest valuethein statement.
the window. (2) Maximum Chromaticityfilter: all the rainbow colours arranged along a horseshoe-shaped
digital
Compression
images :refers Compression
to randomdeals variations
with in techniques
pixel values for malfunctions, or other external factors. - Effect:
information of interest is known to be located, thus limiting It can diagram
100th percentile filter is the maximum filter. The value of
is a convenient space coordinate representation of triangular curve. The various saturated pure spectral
that
reducing
are not thepart storage
of therequired
actual image to save information.
an image Itoris the an significantly
the degrade
search that has image
to be quality by introducing
conducted in seeking high- that all the spectral colours and their mixtures based
the center is replaced by the largest value in the window. on the tri- colours are represented along the perimeter of the curve,
unwanted
bandwidth and to transmit
often unpredictable
it. Particularly interference
in the usesthat of can
the contrast, isolated
information. The pixels.
knowledge- Example:
base alsoImagine
can bea digital
quite stimulus
(3) Median values of Each
filter: the primary
pixel incolours contained
the image by them.
is considered. the corners representing the three primary colours—red,
internet,oritdegrade
distort is very the much necessary
quality of an to compress
image. Two [Link] 7. camera malfunctioning during image capture,
complex, such as an interrelated list of all major leading to a
possible Figure below is a pixels
First neighboring two dimensional
are sorted andrepresentation
original values of hue
of green and blue. -A practical advantage of the chromaticity
Morphological
types of noise areProcessing
Gaussian noise : Morphological
and impulsive noise. processing 1. few pixels
defects in ain materials
the resulting image having
inspection problem extreme
or an values,
image and saturation
the pixel in the X-Y
is replaced by theplane.
medianIf aofthree dimensional
the list. (b) diagram is that, it is possible to determine the result of
deals withNoise:
Gaussian tools for-extracting
Nature: Gaussian image components
noise is a that type are
of either very
database bright high-resolution
containing (salt) or very darkimages
satellite (pepper).
of a representation
Sharpening Spatial is drawn, theIt‘Z’
Filter: axisknown
is also will show relative
as derivative additive mixing of any two or more colour lights by simple
useful in the
statistical representation
noise that follows and description
a Gaussianof shape. or normal 8. Differentiating
region Between
in connection Gaussian Noise and
with change-detection Impulsive
applications brightness of the colour.
filter. The purpose **Wavelength
of the sharpening : Red
spatial filter: is700nm
just the, geometric construction. -The colour diagram contains all
distribution.
SegmentationIn other : Segmentation
words, the intensity procedures values partition
of pixelsan Noise: 1. Distribution: - Gaussian noise follows a Green
opposite of :the smoothing
546.1nm,Blue : Its438.8
spatial filter. main focusnm. in colours of equal brightness “The color gamut of a
affected
image into by itsGaussian
constituent noise partsareorrandomly
objects. In distributed
general, normal distribution, resulting in a smooth variation of pixel on the removal of blurring and highlight the edges. It is monitor can be displayed as a triangular region within the
according
autonomous tosegmentation
a bell-shapedis curve. one of the- most Characteristics:
difficult tasks It values. - Impulsive noise does not follow a specific based on the first and second order derivative. First order CIE diagram, whereas the color gamut of a high quality
appears
in digitalas image a smooth, processing.
continuous A rugged
random segmentation
variation in distribution, and its effect is characterized by isolated, derivative: Must be zero in flat segments. Must be non printing device is displayed as the irregular region inside
procedure across
brightness bringsthethe [Link] a long
- Causes: way toward
Gaussian noise cana extreme pixel values. 2. Appearance: - Gaussian noise zero at the onset of a grey level step. Must be non zero the triangle.” The statement can be reasoned based on the
successful
be introduced solutionduring of the imagingimageproblemsacquisition that process,
require appears as a subtle, grainy texture across the image. - along ramps. First order derivative in 1-D is given by: f' fundamental differences in the color reproduction
objects to be or
transmission, identified
due to sensor individually.
limitations.9.- Representation
Effect: It tends Impulsive noise manifests as distinct, isolated bright and = f(x+1) - f(x). Second order derivative: Must be zero mechanisms of monitors and high-quality printing devices,
andcreate
to Description
a subtle, : Representation
grainy appearance and description
in the image. almost- dark pixels, resembling salt and pepper. 3. Causes: - in flat areas. Must be zero at the onset and end of a ramp. as well as the respective limitations and capabilities of
Example:
always follow Consider the output
a digital ofphotograph
a segmentation takenstage,in low-light
which Gaussian noise is often associated with natural variations Must be zero along ramps. Second order derivative in 1- these technologies. 1. Monitor Color Gamut (Triangular
conditions
usually is raw where pixelthedata,
captured constituting
pixel values eitherare theaffected
boundary by in the imaging process. - Impulsive noise is typically D is given by: f'' = f(x+1) + f(x-1) - 2f(x) .  Example Region): - **Additive Color Mixing:** Monitors typically
random
of a regionvariations,
or all the resulting
points in in the
a slightly
regiongrainy
itself. or Choosing
speckleda caused by sudden, unpredictable events or errors. 4. of Median Filter : It replaces the value at the center by use additive color mixing, where different intensities of
appearance.
representation is [Link] Impulsive
of the solution
Noise (Salt-and-Pepper
for transforming Effect on Images: - Gaussian noise has a less disruptive The median pixel value in the neighborhood, (i.e. the red, green, and blue (RGB) light are combined to produce
raw data- Nature:
Noise): into a form Impulsivesuitable noise, foralso
subsequent
known ascomputer salt-and- impact on image quality. - Impulsive noise can middle element after they are sorted) Median filters are a range of colors. - **RGB Primaries:** The color gamut
processing.
pepper noise, Description
introducesdeals random, withabrupt
extracting
spikes attributes
or dropsthat in significantly degrade image quality, making it more particularly useful in removing impulse noise (also known of a monitor is defined by the RGB primaries, and the entire
result values.
pixel in someIt quantitative
is named after information
the visualofeffect interest or are
it creates, challenging to interpret. 5. Impact: Gaussian noise tends as salt-and-pepper noise). Salt = 255, pepper = 0 gray range of colors that can be displayed results from mixing
basic for differentiating
resembling white and black onespecks
class of objectsthroughout
scattered from another. the to blur the image, while impulsive noise introduces sudden levels. In a 3×3 neighborhood the median is the 5th largest these three primary colors. - **Triangular
image.
10. Object - Characteristics:
recognition : It Recognition
appears as is isolated
the process
bright that
and and noticeable artifacts. 6. Processing: Different value, in a 5x5 neighborhood the 13th largest value, and so Representation:** In the CIE chromaticity diagram, the
dark
assigns
pixels,
a label,making suchitas, more vehicle
noticeable
to an andobject disruptive
based onthan its filtering techniques are often applied to reduce the impact on. color gamut of a monitor is represented as a triangle. This
Gaussian
descriptors. noise.11.- Knowledge
Causes: Impulsive Base : noise Knowledgecan be may causedbe of each type of noise. triangle is formed by connecting the chromaticity
by
as simple
errors as duringdetailing image regions transmission,
of an image where hardware the Gaussian noise is coordinates of the monitor's RGB primaries. 2. Printing
information commonly addressed using smoothing filters, while
d) Explainof or
malfunctions, interest
RGB, other is CMY knownand
external to beHSI
[Link], thus limiting
-color
Effect: It can
models Application- The RGB model is widely used in the g) What is homomorphic filtering? Summarize the Device Color3. **Filtering
components. Gamut (Irregular RegionDomain:**
in the Frequency Inside the-
the search that impulsive noise may require special filters designed to
significantly
emphasizing onhastheir
degrade to berespective
image conducted
quality in seekinghigh-
by introducing
application that
& representation and display of images in electronic systems steps involved in it. Homomorphic filtering is a technique Triangle):
In the frequency- **Subtractive
domain, a Color Mixing:**
filter is applied toHigh-quality
selectively
information. The identify and replace extreme values.
contrast,
[Link] The color model aims to facilitatequite
knowledge
pixels. - base
Example: also
Imaginecan be
a digital
the like computers and televisions. It is also used in used in image processing to enhance the visual printing devices, such as color
adjust the amplitudes of the frequency components. printers, often The use
complex,
camera
specificationssuchof as colors
malfunctioning an interrelated
during
in some imagelist of
standard all way.
capture, major possible
leading to a
Different conventional photography as well. Image scanner which characteristics of an image, particularly when dealing with subtractiveiscolor
objective mixing. This
to enhance involves the combination
the high-frequency components of
defects
few
types of incolor
pixels a materials
in the resulting
models inspection
are imageinhaving
used problem
multiple or an values,
extreme
fields image
like in scans images and converts it to a digital image mostly variations in illumination. It is effective in applications different colorants,
associated with image suchdetails
as inkswhile
or toners, to absorb
suppressing thecertain
low-
either
databasevery
hardware, containingbright high-resolution
in multiple (salt) or very
applications ofsatellitedarkimages
creating (pepper).
of a
animation, supports RGB color. It is used in web graphics. ** where both low-frequency and high-frequency components frequency components related to illumination variations.-
wavelengths of light and produce a variety of colors.
Differentiating
region
etc. In inDigital
connection Between
Image with Gaussian
change-detection
Processing, Noise
the and
applications.
Impulsive
hardware-oriented CMY Model : This model contains the secondary colors. of an image need to be manipulated separately. This **Multiple
The filteringColorants:** Unlike achieved
operation is usually monitors by that use only
multiplying
Noise:
models that are 1. Distribution:
commonly used - Gaussian
are the RGB noise model
followsfora In this model, any secondary color when passed through technique is widely used in applications where images three
the primaries,
frequency printing devices
components use multiple
with a designed filtercolorants,
function.
normal
printersdistribution,
and color monitors. resultingCMY(cyan,
in a smooth magenta,variation of pixel
yellow) white light will not reflect the color from which a suffer from non-uniform illumination, such as satellite sometimes
4. **Inverseincluding specialized inks- or
Fourier Transform:** Thetoners. This
filtered allows
image in
values.
and CMYK(cyan, - Impulsive magenta, noiseyellow,
does black)
not follow models a are
specific
used combination of colors is made. For example- when cyan imaging, medical imaging, and surveillance systems. It for frequency
the a more extensive
domain range of color
is inverse Fourier reproduction.
transformed to-
distribution,
for color printing. and HSI(hue,
its effectsaturation,
is characterizedintensity) bydeals
isolated,
with is illuminated with white light, no red light will be reflected separates and processes the illumination and reflectance **Complex
bring it backPrinting Process:**
to thespatial [Link] color gamut of a
5. **Exponential
extreme
colors aspixelhumans values. interpret. 2. Appearance:
** RGB Model - Gaussian
: noise
In the from the surface which means that the cyan subtracts the components of an image, providing improved visibility of printing device is influenced
Transformation:** - The resultby the properties
obtained of the
from the inverse
appears
RGB model, each color appears in its primary components-
as a subtle, grainy texture across the image. red light from the reflected white light (which itself is details in challenging lighting [Link] main goal of colorantstransform
Fourier used, the is thenprinting
subjectedtechnology, and the
to an exponential
Impulsive
of red, green noise
andmanifests
blue. Thisasmodel distinct, isolated
is based on abright and
Cartesian composed of red, green and white light). The formula given homomorphic filtering is to improve the visibility of details characteristics of the substrate (e.g.,
transformation to revert the logarithmic transformation paper). This
dark pixels, system
coordinate resembling salt and pepper. 3. Causes: -
: -------------------------------------------- is for inter-conversion of RGB and CMY models. in images captured under non-uniform lighting conditions. complexity
applied in theresults
first in an This
step. irregularly
bringsshaped region within
the enhanced image
Gaussian noise is often associated with natural variations **Steps Involved in Homomorphic Filtering:** 1. the RGB
back to itstriangle
originalonintensity
the CIEscale.
diagram. - **Wide and
6. **Normalization
in the imaging process. - Impulsive noise is typically **Logarithmic Transformation:** - The input image is Irregular Gamut:**
(Optional):** - The Printing
enhanced devices
image may canbereproduce
normalizeda
Application- It is used in color
caused by sudden, unpredictable events or errors. 4. initially transformed using a logarithmic function. This broader
to ensureand thatmore
pixelirregular
values are color gamut
within compared
a valid intensityto
printing as it uses colored inks. It is used in most
Effect on Images: - Gaussian noise has a less disruptive logarithmic transformation converts the multiplicative monitors.
range, Thisbetween
usually is because they255
0 and canforgenerate
an 8-bitaimage.
wider range
commercial printing like magazines, books, etc. **HSI
impact on image quality. - Impulsive noise can effects of illumination variations into additive effects. The of colors, including those that fall outside the limits of the
Model (Hue, Saturation, Intensity) It is a very
significantly degrade image quality, making it more RGB color
important and attractive color model because it represents formula for the logarithmic transformation is: Log (I (x, y)) **What dospace.
you mean by a grey scale image ? Grayscale
challenging to interpret. 5. Impact: Gaussian noise tends ,where I (x, y) is the intensity of the pixel at coordinates (x,
the colors the same way as the human eye senses colors. images, a kind of black-and-white or gray monochrome,
to blur the image, while impulsive noise introduces sudden y) in the input image. 2. **Frequency Domain
Hue is a color component that describes a pure color (pure are composed exclusively of shades of gray. The contrast
and noticeable artifacts. 6. Processing: Different Representation:** - The logarithmically transformed
yellow, orange or red) Saturation component represents ranges from black at the weakest intensity to white at the
filtering techniques are often applied to reduce the impact image is then converted into the frequency domain using
the measure of the degree to which color is mixed with strongest.
of each type of noise. Gaussian noise is commonly the Fourier transform. This allows for the separation of the
white color. 0 degree – Red 120 degree – Green 240
addressed using smoothing filters, while impulsive noise image into its low-frequency and high-frequency
degree – Blue 60 degree – Yellow 300 degree – Magenta
may require special filters designed to identify and replace components. 3. **Filtering in the Frequency Domain:** -
Intensity : Range is [0, 1] 0 means white
extreme values. In the frequency domain, a filter is applied to selectively
h) Multiresolution Techniques : Multiresolution Hierarchical Models : A model in which a pixellated, adjust the amplitudes
j) Discuss the concept, of the frequency
structure andcomponents.
applicationsThe of - **Usage:** - Laplacian pyramids are used in image
Techniques span an exceptionally broad range of finest-scale random field is explicitly represented using a objective is to enhanceAn
image pyramids. theimage
high-frequency
pyramid is acomponents
multi-scale compression, coding, and edge detection. Applications of
algorithms, models, methods, and concepts. Central to the set of random fields over scales. In many cases the multi- associated with image details while suppressing
representation of an image that consists of a series of the low- Image Pyramids: 1. **Image Blending:** - Image
multiresolution approach is to somehow express short- scale model may be simpler, using principles of Markov frequency components
images at different related to
resolutions. illumination
Each level of the variations.
pyramid pyramids are used for seamless blending of two images at
range, mid-range, and long-range relationships explicitly. decomposition to decouple the problem into pieces. A The filtering
represents theoperation
image atisa usually
differentachieved by multiplying
scale, with the base of different scales, providing a smooth transition. 2. **Image
The main reasons for a multiresolution approach is one of: multi-scale model allows different models to be asserted at the frequency
pyramid being components withimage.
the original a designed
Imagefilter function.
pyramids are Compression:** - Pyramids, especially Laplacian
improving performance, by capturing long-range different scales, usually simpler or more meaningful than 4. **Inverse
widely used Fourier Transform:**
in computer vision and- image
The filtered imagefor
processing in pyramids, are used in image compression algorithms to
phenomena that would otherwise not be utilized. having a single-scale model which needs to assert all of the the
tasksfrequency
such as domain is inverseimage
image blending, Fourier transformed
matching, to
feature efficiently represent and reconstruct images. 3. **Feature
reducing computational complexity, by allowing various sccale-dependent behaviours simultaneously. bring it back
detection, to thespatial
and image domain. Structure
segmentation. 5. **Exponential
of Image Matching:** - In computer vision, image pyramids are
algorithms to work on both fine and coarse scales, rather Hierarchical Algorithms : Even if there is no explicitly Transformation:**
Pyramids: There are - The
tworesult
main obtained from the
types of image inverse
pyramids: employed for feature matching across different scales,
than waiting for local pixel-level operations to converge at hierarchical model, it is possible for the processing Fourier
Gaussiantransform
pyramid and is then subjected
Laplacian to an [Link]
pyramid. Gaussian improving robustness. 4. **Image Registration:** -
large scales. improving numerical robustness (reducing algorithm to be hierarchical. Best known examples include transformation to revert the logarithmic
Pyramid: - **Construction:** transformation
- The Gaussian pyramid Image pyramids help in achieving scale-invariant image
problem conditioning), whereby a multiresolution multigrid methods, whereby a single-scale linear system is applied
is built byin repeatedly
the first step. This brings
applying the enhanced
a Gaussian smoothingimage filter registration, aligning images captured at different
transformation is essentially an algebraic pre-conditioner. solved by casting the problem onto a hierarchy, and back
to theto its original
original imageintensity
and then scale. 6. **Normalization
down-sampling it. - Each resolutions. 5. **Object Recognition:** - Multi-scale
simplifying the algorithm, by making accessible long- wavelet methods in image processing, whereby the image (Optional):**
level of the pyramid - Theisenhanced
obtained by image may bethe
convolving normalized
previous representations provided by pyramids are useful in object
range features that might, in some problems, be much is transferred into a set of multiscale coefficients in the to ensure
level with that pixel values
a Gaussian kernelare within its
to reduce a valid intensity-
resolution. recognition tasks where objects may appear at different
easier to work with than pixel-level features improving wavelet domain, in which certain operations (like image range, usually between
**Properties:** 0 andlevels
- Higher 255 for an 8-bit
of the pyramidimage.
represent scales. 6. **Mipmap Texturing:** - In computer
intuition, by modeling or analyzing the problem over compression or image denoising) are relatively simple. lower-resolution images. - The base of the pyramid is the graphics, pyramids are used for mipmapping, which is a
multiple scales, getting deeper insights into the original image. - **Usage:** - Gaussian pyramids are technique for texture mapping to reduce aliasing artifacts.
phenomenon at hand. Although there are, for sure, many ** What is Gamma correction? Why such correction is commonly used for image smoothing, down-sampling, and 7. **Image Segmentation:** - Pyramids aid in multi-scale
mutiresolution approaches and algorithms which have been required? Gamma correction is responsible for image blending. 2. Laplacian Pyramid: - image segmentation by allowing segmentation algorithms
proposed, broadly these fall into a few groups: Wavelet performing nonlinear methods on the pixels of the input **Construction:** - The Laplacian pyramid is to operate at different resolutions. 8. **Pyramid-Based
Methods : Problems in which a wavelet transform is used image and thereby remodeling the saturation of the image. constructed by taking the difference between each level of Optical Flow:** - In motion analysis, pyramid structures
to decompose an image or video into multiple scales, very It is necessary to maintain the stable gamma value, i.e., it the Gaussian pyramid and its expanded version are utilized for pyramid-based optical flow algorithms to
commonly for image/video denoising, or for feeding the should neither be too minimum or maximum. (upsampled). - It is essentially the high-pass information estimate motion at different scales. Image pyramids
coefficients at multiple scales into a classifier for image at different scales. - **Properties:** - Each level are versatile tools that provide a flexible representation of
classification and segmentation. Hierarchical Models : represents the details or high-frequency components. - images at multiple scales, enabling various applications.
A model in which a pixellated, finest-scale random field is The base of the pyramid contains the low-frequency in computer vision, image processing, and
value, in a 5x5 neighborhood the 13th largest value, and so a) What is city-block distance? Explain with a suitable 10. Object recognition : Recognition is the process that In these equations, represents the complex amplitude of the
on. Let’s, a 3×3 neighborhood has gray levels (10, 20, 0, example. The City block distance between two assigns a label, such as, vehicle to an object based on its frequency component at spatial frequency indices and . The
20, 255, 20, 20, 25, 15) These values are sorted as points, a and b, with k dimensions is calculated as: descriptors. 11. Knowledge Base : Knowledge may be indices and iterate over the spatial coordinates of the 4x4
(0,10,15,20,20,20,20,25,255), which results in a median of The City block distance is always greater than or equal to as simple as detailing regions of an image where the sub-image. Two-Dimensional DCT Pair: The 2D DCT
20 that replaces the original pixel value 255 (salt noise). zero. The measurement would be zero for identical points information of interest is known to be located, thus limiting pair for a 4x4 sub-image can be expressed as follows:
and high for points that show little similarity. The figure the search that has to be conducted in seeking that Forward 2D DCT: ---------------------------------------------
Consider the following 5*5 below shows an example of two points called a and b. information. The knowledge base also can be quite
[Link] a 3x3 median Each point is described by five values. The dotted lines in complex, such as an interrelated list of all major possible
filter on the shaded pixels, the figure are the distances (a1-b1), (a2-b2), (a3-b3), (a4-b4) defects in a materials inspection problem or an image Inverse 2D DCT:-----------------------------------------------
and write the filtered mage. and (a5-b5) which are entered in the equation above. database containing high-resolution satellite images of a
Solution :- region in connection with change-detection applications.
------------------------------- In these equations, represents the DCT coefficients, and
B) Write down the expressions for two dimensional and are the scaling factors defined as:
In most cases, this DFT pair and DCT pair for a 4×4 sub-image
distance measure yields The Discrete Fourier Transform (DFT) and Discrete
results similar to the Euclidean distance. Note, however, Cosine Transform (DCT) are mathematical
that with City block distance, the effect of a large transformations commonly used in image processing and
difference in a single dimension is dampened (since the compression. For a 4x4 sub-image, let's denote the pixel
distances are not squared). The name City block values as f(x, y) where and are the spatial coordinates.
Two-Dimensional DFT Pair: The 2D DFT pair for a These equations represent the forward and inverse
distance is explained if two points in the xy-plane is
4x4 sub-image can be expressed as follows: i) Forward 2D transformations between the spatial domain and frequency
considered. The shortest distance between the two points
DFT: domain using the DFT and DCT, respectively, for a 4x4
is along the hypotenuse, which is the Euclidean distance.
sub-image. ---------! ** Degradation comes in many
The City block distance is instead calculated as the
forms such as motion blur, noise, and camera misfocus. In
distance in x plus the distance in y, which is similar to the
ii) Inverse 2D DFT:_______________________________ cases like motion blur, it is possible to come up with an
way you move in a city where you have to move around
very good estimate of the actual blurring function and
( Filtered Image ) the buildings instead of going straight through______
"undo" the blur to restore the original image.

triangle is formed by connecting the chromaticity c) Histogram equalization is a widely used contrast- e) Find the output after applying Median filter to the applied to non-repetitive data. i) Exemplification:
coordinates of the monitor's RGB primaries. 2. Printing enhancement technique in image processing because of its sub-image. What type of noise is removed by this filter? Original Data: Consider the following sequence of pixel
Device Color Gamut (Irregular Region Inside the high efficiency and simplicity. It is one of the sophisticated camera
Explainmalfunctioning
-The sorted during
values image
in the capture, leading to a
3x3 neighborhood values in an image: 1 1 1 3 3 3 3 0 0 2___Run-
Triangle): - **Subtractive Color Mixing:** High-quality methods for modifying the dynamic range and contrast of few
werepixels
[2, 2, in 3, the
4, 4,resulting image
5, 4, 5, 6]. -Thehaving
median extreme values,
value (middle Length Encoding: Applying run-length encoding to the
printing devices, such as color printers, often use an image by altering that image such that its intensity either
value) invery brightlist(salt)
this sorted or very thedark
is 4.-Therefore, center(pepper).
pixel in original data, we get: (1, 3) (3, 4) (0, 2) 2 _____In
subtractive color mixing. This involves the combination of histogram has the desired shape. It can be classified into Differentiating Between Gaussian Noise
the original sub-image, which was initially 5, is replaced and Impulsive the encoded sequence, each pair represents a run of
different colorants, such as inks or toners, to absorb certain two branches as per the transformation function is used. Noise: 1. Distribution:
with 4 after applying the median - Gaussian
filter. -The noise follows
median filtera consecutive identical elements, where the first value is the
wavelengths of light and produce a variety of colors. - Global histogram equalization (GHE) GHE is very normal distribution,
is a robust choice resulting in a smooth
for removing impulsevariation
noiseof while
pixel pixel value, and the second value is the count of
**Multiple Colorants:** Unlike monitors that use only simple and fast, but its contrast enhancement power is low. values. - Impulsive noise does
preserving the details and edges in an image. Type ofnot follow a specific consecutive occurrences. ii) Contrast: 1. Redundancy
three primaries, printing devices use multiple colorants, Here the histogram of the whole input image is used to distribution,
Noise Removed: and Theits effect
medianisfilter
characterized
is particularlyby effective
isolated, Reduction: - **Exemplification:** In the original data,
sometimes including specialized inks or toners. This allows compute the histogram transformation function. As a extreme
at removingpixelimpulse
values. noise, 2. also
Appearance:
known as-salt-and-pepper
Gaussian noise there is redundancy in the consecutive occurrences of the
for a more extensive range of color reproduction. - result, the dynamic range of the image histogram is appears as a subtle,
noise. Impulse noise grainy texture
introduces acrossbright
isolated the image.
(salt) and- same pixel values. Run-length encoding reduces this
**Complex Printing Process:** The color gamut of a flattened and stretched. The overall contrast is improved. Impulsive
dark (pepper) noise manifests
pixels as distinct,
in the image. In this isolated bright
case, the and
median redundancy by representing each run with a single value
printing device is influenced by the properties of the Local histogram equalization (LHE) LHE can enhance dark
filter pixels,
replaces resembling
each pixelsalt withandthepepper.
median value 3. Causes:
of its 3x3- and its count. - **Contrast:** Redundancy reduction is
colorants used, the printing technology, and the the overall contrast more effectively. One of the drawbacks Gaussian
neighborhood, noisewhich
is often associated
helps eliminatewith natural of
the impact variations
extreme achieved by avoiding the repetition of the same
characteristics of the substrate. This complexity results in of histogram equalization is that it can change the mean in the imaging
values caused by process. - Impulsive noise is typically
noise.---------------------------------------- information and storing it more compactly.2. Efficiency in
an irregularly shaped region within the RGB triangle on the brightness of an image significantly as a consequence of caused by sudden, unpredictable events
f) Exemplify and contrast the redundancy reduction in or errors. 4. Repetitive Patterns: - **Exemplification:** Run-length
CIE diagram. - **Wide and Irregular Gamut:** Printing histogram flattening and sometimes this is not a desirable Effect on Images:
run length coding.- Gaussian
Run-length noisecoding
has a less is adisruptive
form of coding is highly efficient when dealing with images or
devices can reproduce a broader and more irregular color property when preserving the original mean brightness of a impact on image
lossless data compressionquality. - Impulsive
where sequences noise can
of consecutive sequences where consecutive identical elements occur
gamut compared to monitors. This is because they can given image is necessary. Bi-Histogram Equalization was significantly
identical elements degrade (runs)image quality,with
are replaced making
a single it value
more frequently. - **Contrast:** In scenarios where there are
generate a wider range of colors, including those that fall proposed to overcome this problem. Steps Involved Get challenging
and its count. toThe
interpret.
goal is [Link]
Impact: Gaussian noise
redundancy in thetends
data long runs of identical values, run-length coding can
outside the limits of the RGB color space. In summary, the input image,Generate the histogram for the image,Find to
byblur the image,repetitive
representing while impulsive
patternsnoise moreintroduces
efficiently. sudden
run- significantly reduce the amount of storage needed to
the triangular region within the CIE diagram represents the the local minima of the image,Divide the histogram based and noticeable artifacts. 6. Processing:
length coding is a simple yet effective method for reducing Different represent the data.3. Lossless Compression: -
color gamut of a monitor, which is limited by the RGB on the local minima,Have the specific gray levels for each filtering
redundancy techniques
in data, are often applied when
particularly to reduce the impact
dealing with **Exemplification:** Run-length coding is a lossless
primaries and additive color mixing. On the other hand, the partition of the histogram,Apply the histogram equalization of each type
sequences of noise.
or images thatGaussian
contain runs noise ofis consecutive
commonly compression method, meaning that the original data can be
irregular region inside the triangle represents the color on each partition addressed
identical elements. It is efficient in scenarios wherenoise
using smoothing filters, while impulsive such perfectly reconstructed from the encoded data. -
gamut of a high-quality printing device. may
runs require special filters
are prevalent, but itsdesigned
effectivenessto identify and replace
diminishes when **Contrast:** Lossless compression methods ensure that
, which can achieve a broader and more extreme
applied values.
to non-repetitive data. i) Exemplification: no information is lost during the encoding and decoding
irregular range of colors through subtractive color mixing Original Data: Consider the following sequence of pixel processes.
with multiple
i) Discuss thecolorants. The irregular
basic operations involved shapeinreflects
geometric the pixel values at the new locations based on the surrounding values in :an
Intensity image:
Range 1 10 means
is [0, 1] 1 white
3 3 13means 3 black
0 0 An Expert System is a computer program that is designed
complex and nuanced
transformation color reproduction
of an image. Geometric capabilities
transformation of pixel values. Applications of Geometric 2_____Run-Length
The formula for conversionEncoding: of RGB Applying
to HSIrun-length
is quite to solve complex problems and to provide decision-making
printing
is a fundamental technique used in image processing that technologies. Transformation : Geometric transformation is a versatile encoding to the original data, we get: (1, 3)
complicated as compared to other color models. HSI Model (3, 4) (0, ability like a human expert. It performs this by extracting
involves manipulating the spatial arrangement of pixels in technique that finds a wide range of applications in image 2) 2 _____In the encoded sequence, each pair represents knowledge from its knowledge base using the reasoning
an image. It is used to modify the geometric properties of processing. Some of the common applications of geometric a run of consecutive identical elements, where the first and inference rules according to the user queries.
an image, such as its size, shape, position, and orientation. transformation are as follows: 1. Image Registration: value is the pixel value, and the second value is the count Capabilities of the Expert System: [Link]: It is capable
The following are some of the basic concepts of geometric Image registration is the process of aligning two or more of consecutive occurrences. ii) Contrast: 1. of advising the human being for the query of any domain
transformation in image processing: Basic Concepts of images to enable comparison or analysis. Geometric Redundancy Reduction: - **Exemplification:** In the from the particular ES. [Link] decision-making
Geometric Transformation: 1. Transformation transformation is used to align the images by modifying original data, there is redundancy in the consecutive capabilities: It provides the capability of decision making
Functions: Transformation functions are mathematical their position, orientation, or scale. 2. Image Rectification: occurrences of the same pixel values. Run-length encoding in any domain, such as for making any financial decision,
functions used to modify the geometric properties of an Image rectification is the process of removing distortions reduces this redundancy by representing each run with a decisions in medical science, etc. [Link] a device:
The angle from the red
image. These functions map the coordinates of each pixel from an image caused by the projection of a three- single value and its count. - **Contrast:** Redundancy It is capable of demonstrating any new products such as its
axis gives the Hue, the length of the vector is the saturation
in an image to new coordinates based on a specified dimensional scene onto a two-dimensional image plane. reduction is achieved by avoiding the repetition of the same features, specifications, how to use that product, etc.
and the intensity is given by the position of the plane on the
transformation rule. Some of the commonly used Geometric transformation is used to correct the perspective information and storing it more compactly.2. Efficiency in [Link]-solving: It has problem-solving capabilities.
vertical intensity [Link] biggest application of this model
transformation functions include scaling, rotation, distortion by transforming the image into a new coordinate Repetitive Patterns: - **Exemplification:** Run-length [Link] a problem: It is also capable of providing a
is that it represents colors similarly to how the human eye
translation, and shearing. 2. Coordinate System: A system. 3. Image Resizing: Image resizing is the process coding is highly efficient when dealing with images or detailed description of an input problem. 6. Interpreting the
senses colors.------------------------------------------------------
coordinate system is a reference system used to define the of changing the size of an image while preserving its aspect sequences where consecutive identical elements occur input: It is capable of interpreting the input given by the
spatial location of pixels in an image. In digital image ratio. Geometric transformation is used to scale the image frequently. - **Contrast:** In scenarios where there are user. 7. Predicting results: It can be used for the prediction
**What
long runs is image restoration?
of identical values,How it is different
run-length codingfrom can
processing, the most commonly used coordinate system is by a factor in the x and y directions. 4. Image Rotation: of a result. 8. Diagnosis: An ES designed for the medical
contrast enhancement?
significantly reduce the amountimage restoration
of storageis focused
needed on to
the Cartesian coordinate system, which uses two Image rotation is the process of rotating an image by a field is capable of diagnosing a disease without using
recovering the original scene from a
represent the data.3. Lossless Compression:degraded image, while -
perpendicular axes (x and y) to represent the horizontal and specified angle. Geometric transformation is used to multiple components as it already contains various inbuilt
contrast enhancement is concerned with
**Exemplification:** Run-length coding is a lossless improving the
vertical positions of pixels in an image. 3. Interpolation: perform the rotation by mapping the original pixel medical tools. Advantages: These systems are highly
visual appearance
compression method,andmeaning
emphasizing certain
that the features
original in an
data can be
Interpolation is the process of estimating the pixel values positions to new positions based on the rotation angle. reproducible. [Link] can be used for risky places where the
image.
perfectlyBoth processes from
reconstructed play crucial
the encodedroles in image-
data.
of an image at locations that are not explicitly defined. This human presence is not safe. [Link] possibilities are less if
processing,
**Contrast:** addressing
Lossless different aspects
compression of image
methods quality
ensure that
is necessary when transforming an image since the new the KB contains correct knowledge. 4. They provide a very
improvement.
no information is lost during the encoding and decoding
coordinates may not coincide with the original pixel high speed to respond to a particular query .
positions. Interpolation algorithms are used to estimate the processes.
pixel values at the new locations based on the surrounding
pixel values. Applications of Geometric
Transformation
K : Geometric
) What is the objective transformation
of image compression? is a versatile and devices. Some widely used image compression L ) Discuss about wavelet transforms in one and two coefficient [1, -1, 1, -1]. (2D) Wavelet Transforms:
technique that finds a wide range of
Explain with an example of a high definition (HD) applications in image standards include: 1. **JPEG (Joint Photographic Experts dimensions. Wavelet transforms are mathematical In 2D wavelet transforms, the process is extended to two
processing.
digital videoSome [Link]
the common
are theapplications
various Image of geometric Group):** - Widely used for compressing still images. - techniques used for analyzing and processing signals, dimensions, making it suitable for the analysis of images
transformationStandards?
Compression are as follows:[Link] Image objective
Registration: of Lossy compression algorithm. - Suited for natural images images, and other types of data. They decompose a signal and other 2D data. The most common 2D wavelet
Image compression
image registration isis the processtheofsize
to reduce aligning two images
of digital or more and photographs. 2. **JPEG 2000:** - An improvement into components at different scales and frequencies, transform is the 2D Discrete Wavelet Transform (2D
images to enable comparison or analysis.
or videos while maintaining an acceptable level of quality. Geometric over the original JPEG standard. - Utilizes wavelet-based making them particularly effective for representing both DWT). The process involves the following steps: 1.
transformation
Image compression is used to alignfor
is essential theefficient
images storage,
by modifying compression. - Supports both lossless and lossy high and low-frequency information. Wavelet transforms **Decomposition (Analysis):** - The image is
their position,and
transmission, orientation,
processingor scale.
of digital2. Image
visualRectification:
information. compression.3. **PNG (Portable Network Graphics):** - are widely used in various applications, including image decomposed along rows and columns using low-pass and
The compression process reduces redundancy indistortions
Image rectification is the process of removing the data, Widely used for lossless compression of images. - compression, denoising, feature extraction, and signal high-pass filters. - Subsampling is performed in both
from an image caused by the projection
eliminating unnecessary information while retaining of a three-
the Particularly suitable for images with transparency.4. **GIF processing. There are two main types of wavelet dimensions. 2. **Reconstruction (Synthesis):** - The
dimensional
essential sceneof onto
features a two-dimensional
the image or video. Example [Link] (Graphics Interchange Format):** - Commonly used for transforms: one-dimensional (1D) wavelet transforms and decomposition process can be inverted to reconstruct the
Geometric
Digital Videotransformation
Data: Consider is used to correct the perspective
a high-definition (HD) simple graphics, logos, and animations. - Supports lossless two-dimensional (2D) wavelet transforms. (1D) Wavelet original image. - Like in 1D, multi-resolution analysis is
distortion by transforming the image into
digital video that captures a scene with a resolution of a new coordinate compression but has a limited color palette.5. **WebP:** Transforms: In 1D wavelet transforms, the signal is possible with different levels of detail. **Example:**
system. 3. Image
1920x1080 Resizing:
pixels (Full HD). Image
Each frameresizing is the
of the videoprocess - Developed by Google, a modern image format. - decomposed into approximation and detail coefficients at Consider a 2D image represented by a matrix of pixel
of changing
consists the size of
of millions of pixels,
an image andwhile preserving
the video comprisesits aspect
a Employs both lossy and lossless compression. - Designed different scales. The most commonly used 1D wavelet values. The 2D DWT decomposition will yield
sequence of these frames. Without compression, the image
ratio. Geometric transformation is used to scale the size for better compression efficiency compared to JPEG.6. transform is the Discrete Wavelet Transform (DWT). The approximation and detail coefficients for both rows and
by the
of a factor
video in thecan
data x and y directions. large,
be prohibitively 4. Image
making Rotation:
it **H.264 (Advanced Video Coding, AVC):** - Widely process involves the following steps: 1. **Decomposition columns. Properties and Applications: -
Image rotation
challenging is thetransmit,
to store, processorofstream
rotating an imageFor
efficiently. by a used for video compression. - Supports high-quality (Analysis):** - The input signal is successively filtered **Multiresolution Analysis:** Wavelet transforms provide
specified A
example:- angle. Geometric
10-second Full HD transformation
video at 30 frames is usedper to compression with relatively low bit rates. - Commonly with a low-pass filter (for approximation) and a high-pass a multi-resolution representation of signals and images,
perform the rotation by mapping
second without compression would require a substantialthe original pixel used in streaming services, video conferencing, and digital filter (for details). - Subsampling is performed to reduce capturing both fine and coarse details. - **Image
positionsoftostorage
amount new positions
space. - based on the rotationthe
With compression, angle.
video television.7. **H.265 (High Efficiency Video Coding, the resolution. 2. **Reconstruction (Synthesis):** - The Compression:** Wavelet transforms are widely used in
file size can be significantly reduced while maintaining HEVC):** - An improvement over H.264, offering higher decomposition process can be inverted to reconstruct the image compression algorithms such as JPEG2000, where
visual quality, making it more practical for storage, compression efficiency. - Used for 4K video streaming original signal. - The signal can be reconstructed with they efficiently represent image content with different
streaming, and transmission over networks. **Various and ultra-high-definition (UHD) content. different levels of detail, allowing for multi-resolution levels of detail. **Denoising:** Wavelet thresholding
Image Compression Standards:** Several image analysis. **Example:** Consider a 1D signal [4, 6, 10, 12, techniques are applied for denoising signals and images by
compression standards have been developed to ensure 8, 7, 4, 2]. The DWT decomposition at one level might removing or suppressing high-frequency noise
interoperability and compatibility across different systems yield an approximation coefficient [7, 11, 9, 3] and a detail components.
and devices. Some widely used image compression coefficient [1, -1, 1, -1]. (2D) Wavelet Transforms: - **Feature Extraction:** Wavelet