Term Work

MASTER IN COMPUTER APPLICATION

YEAR 2024-2025

By
RANJIT CHANDRAKANT JADHAV (41648)
SEAT NO :
SEM II

Institute of Distance and Open learning

Vidya Nagari, Kalina, Santacruz East – 400098
University of Mumbai

PCP CENTER
[Shree Ram College of Commerce, Bhandup]
SUB : AI & ML
Module 1: Introduction to AI
1. What is Artificial Intelligence (AI)?

○ AI involves creating systems that simulate human intelligence processes.

○ These processes include learning, reasoning, and self-correction.

○ AI applications span expert systems, natural language processing, and robotics.

○ The goal is to develop systems capable of performing tasks requiring human

intelligence.

2. What are intelligent agents?

○ Intelligent agents perceive their environment and take actions to achieve specific
goals.

○ They can be simple reflex agents or more complex ones with learning
capabilities.

○ These agents use sensors to gather data and actuators to perform actions.

○ They aim to maximize their chances of success through autonomous

decision-making.

3. What is the PEAS representation?

○ PEAS stands for Performance measure, Environment, Actuators, and Sensors.

○ It's a framework used to specify the structure of an intelligent agent.

○ Performance measure defines the success criteria of the agent's actions.

○ Environment refers to the context in which the agent operates, while actuators
and sensors are the mechanisms through which the agent interacts with and
perceives its environment.

4. What is propositional logic?

○ Propositional logic deals with propositions that can either be true or false.

○ It's used in AI for knowledge representation and reasoning.

○ Logical connectives like AND, OR, and NOT are used to build complex
expressions.

○ It forms the foundation for more complex logical systems in AI.

Module 2: Search Strategies

1. What is Breadth-First Search (BFS)?

○ BFS is an uninformed search algorithm that explores all nodes at the present
depth level before moving on to nodes at the next depth level.

○ It guarantees the shortest path in an unweighted graph.

○ BFS uses a queue data structure to keep track of nodes to visit.

○ It's particularly useful in scenarios where the shortest path is required.

2. What is Depth-First Search (DFS)?

○ DFS is an uninformed search algorithm that explores as far down a branch as

possible before backtracking.

○ It uses a stack data structure to manage the nodes.

○ DFS can get stuck in deep or infinite branches if not properly managed.

○ It's memory-efficient compared to BFS but may not find the shortest path.
3. What is the A algorithm?*

○ A* is a heuristic search algorithm that finds the shortest path from a start node to
a goal node.

○ It combines the cost to reach the node and the estimated cost to the goal.

○ A* uses a priority queue to explore nodes with the lowest total cost.

○ It's widely used in pathfinding and graph traversal.

4. What is Hill Climbing?

○ Hill Climbing is a heuristic search algorithm that continuously moves towards the
direction of increasing value to find the peak or goal state.

○ It can get stuck in local maxima, failing to find the global maximum.

○ Variants like Steepest-Ascent Hill Climbing and Stochastic Hill Climbing address
some of its limitations.

○ It's used in optimization problems where the solution space is large.

Module 3: Artificial Neural Networks

1. What is a Perceptron?

○ A Perceptron is a type of artificial neural network that performs binary

classification.

○ It consists of a single layer of input nodes connected to an output node.

○ The output is determined by a weighted sum of inputs passed through an

activation function.

○ It's the simplest form of a neural network and serves as a building block for more
complex networks.

2. What is Backpropagation?
○ Backpropagation is a supervised learning algorithm used for training neural
networks.

○ It calculates the gradient of the loss function with respect to each weight by the
chain rule.

○ The weights are then updated to minimize the error.

○ Backpropagation is essential for training deep neural networks.

3. What is Gradient Descent?

○ Gradient Descent is an optimization algorithm used to minimize the loss function

in machine learning models.

○ It involves taking repeated steps in the opposite direction of the gradient of the
function at the current point.

○ Variants include Stochastic Gradient Descent and Mini-batch Gradient Descent.

○ It's widely used in training neural networks and other machine learning models.

4. What are Multilayer Perceptrons (MLP)?

○ MLPs are feedforward artificial neural networks with multiple layers of neurons.

○ They can model complex relationships in data and are capable of learning
non-linear functions.

○ MLPs are trained using algorithms like backpropagation.

○ They are foundational to deep learning and have applications in various domains.

Module 4: Introduction to Machine Learning

1. What is Supervised Learning?

○ Supervised Learning is a type of machine learning where the model is trained on

labeled data.
○ The algorithm learns to map input data to the correct output labels.

○ Common algorithms include Naive Bayes, k-Nearest Neighbors, and Decision

Trees.

○ It's used in classification and regression tasks.

2. What is Unsupervised Learning?

○ Unsupervised Learning involves training a model on data without labeled outputs.

○ The goal is to identify underlying patterns or groupings in the data.

○ Common techniques include clustering and association analysis.

○ It's used in exploratory data analysis and dimensionality reduction.

3. What is k-Nearest Neighbors (k-NN)?

○ k-NN is a simple, non-parametric algorithm used for classification and regression.

○ It classifies a data point based on the majority class of its k nearest neighbors.

○ The distance metric, such as Euclidean distance, is used to find the nearest
neighbors.

○ k-NN is intuitive and effective for small datasets.

4. What is the Apriori Algorithm?

○ The Apriori Algorithm is used for mining frequent itemsets and learning
association rules in transactional databases.

○ It operates on the principle that a subset of a frequent itemset must also be a

frequent itemset.

○ Apriori uses a breadth-first search strategy to count itemset frequencies.

○ It's widely used in market basket analysis.

Module 5: Forecasting and Learning Theory
1. What is Logistic Regression?

○ Logistic Regression is a statistical model used for binary classification problems.

○ It predicts the probability that an instance belongs to a particular class using the
logistic function.

○ The output is a value between 0 and 1, representing the probability.

○ It's widely used in fields like medicine and finance.

2. What is Random Forest?

○ Random Forest is an ensemble learning method that creates multiple decision

trees.

○ Each tree is trained on a random subset of the data, and their outputs are
averaged for regression or voted on for classification.

○ It reduces overfitting and improves generalization.

○ Random Forest is robust and handles large datasets well.

3. What is the Bias-Variance Tradeoff?

○ The bias-variance tradeoff refers to the balance between bias (error due to overly
simplistic models) and variance (error due to overly complex models).

○ High bias can lead to underfitting, while high variance can lead to overfitting.

○ The goal is to find a model with both low bias and low variance.

○ This balance is crucial for building effective machine learning models.

4. What is Expectation-Maximization (EM)?

○ The EM algorithm is an iterative method used for parameter estimation in models

with latent variables.

○ It alternates between estimating the missing data (E-step) and optimizing the
parameters (M-step).
○ EM is used in clustering and density estimation.

○ It's particularly useful when data has missing or incomplete information.

Module 6: Kernel Machines & Ensemble Methods

1. What is a Support Vector Machine (SVM)?

○ SVM is a supervised learning algorithm used for classification and regression

tasks.

○ It finds the hyperplane that best separates data points of different classes.

○ SVM can handle both linear and nonlinear classification using kernel functions.

○ It is effective in high-dimensional spaces and for cases where the number of

dimensions exceeds the number of samples.

2. What is the Kernel Trick?

○ The kernel trick is a method used in SVMs to enable learning in

higher-dimensional spaces without explicitly computing the coordinates in that
space.

○ It uses a kernel function to compute the inner product between two points in the
feature space.

○ This allows SVMs to fit the maximum-margin hyperplane in a transformed feature

space.

○ Common kernel functions include linear, polynomial, and radial basis function
(RBF) kernels.

3. What are Ensemble Methods?

○ Ensemble methods combine multiple machine learning models to improve

performance.

○ They aim to reduce variance (bagging), bias (boosting), or improve predictions

(stacking).
○ Examples include Random Forests, AdaBoost, and Gradient Boosting Machines.

○ Ensemble methods often outperform individual models by leveraging their

collective strengths.

4. What is AdaBoost?

○ AdaBoost (Adaptive Boosting) is an ensemble technique that combines multiple

weak classifiers to create a strong classifier.

○ It adjusts the weights of incorrectly classified instances so that subsequent

classifiers focus more on difficult cases.

○ AdaBoost is sensitive to noisy data and outliers.

○ It is effective for binary classification tasks and can be adapted for multi-class
problems.

Module 7: Dimensionality Reduction

1. What is Dimensionality Reduction?

○ Dimensionality reduction is the process of reducing the number of input variables

in a dataset.

○ It helps in simplifying models, reducing computational cost, and mitigating the

curse of dimensionality.

○ Techniques include feature selection and feature extraction.

○ It is particularly useful in preprocessing high-dimensional data.

2. What is Principal Component Analysis (PCA)?

○ PCA is a statistical technique used to identify the directions (principal

components) along which the variance of the data is maximized.
○ It transforms the data into a new coordinate system, reducing dimensionality
while retaining most of the variance.

○ PCA is widely used for exploratory data analysis and preprocessing before
machine learning.

○ It assumes linear relationships among variables and is sensitive to scaling.

3. What is Linear Discriminant Analysis (LDA)?

○ LDA is a dimensionality reduction technique that focuses on maximizing the

separability among known categories.

○ It projects data onto a lower-dimensional space with a goal of preserving class

discriminatory information.

○ LDA is supervised, unlike PCA, which is unsupervised.

○ It is commonly used in pattern and face recognition tasks.

4. What is Feature Selection?

○ Feature selection involves selecting a subset of relevant features for use in

model construction.

○ It helps in improving model performance, reducing overfitting, and decreasing

computational cost.

○ Methods include filter, wrapper, and embedded techniques.

○ Feature selection is an essential step in the machine learning pipeline, especially

for high-dimensional datasets.
SUB: Natural Language Processing

What is NLP? Discuss the different stages in NLP.

Natural Language Processing (NLP) is a branch of artificial intelligence that
enables computers to understand, interpret, and respond to human language.
It combines linguistics and computer science to process text and speech data
effectively.
The key stages in NLP are:
1. Tokenization: Splitting text into words, phrases, or sentences.
2. Stop Word Removal: Filtering out common words like “and”, “is”, which
carry less meaning.
3. Stemming and Lemmatization: Reducing words to their base or root
form (e.g., "running" to "run").
4. Part-of-Speech Tagging: Identifying word categories (noun, verb, etc.).
5. Named Entity Recognition (NER): Detecting names of people, places,
organizations, etc.
6. Parsing: Analyzing grammatical structure.
7. Sentiment Analysis: Understanding the emotional tone of the text.
These stages help in tasks like translation, chatbots, and voice recognition by
making human language comprehensible to machines.

Explain with examples Inflectional morphology &

Derivational morphology
Inflectional morphology and derivational morphology are branches of
morphology, the study of word formation.
Inflectional morphology modifies a word to express grammatical features like
tense, number, or case, without changing its core meaning or word class. For
example:
 Talk → Talked (past tense)
 Cat → Cats (plural)
These changes show grammar but keep the word’s basic meaning and
function.
Derivational morphology, on the other hand, creates new words by adding
prefixes or suffixes, often changing the word's meaning or category. For
example:
 Happy (adjective) → Happiness (noun)
 Teach (verb) → Teacher (noun)
Here, derivational affixes form new words with new meanings.
In summary, inflectional morphology adjusts words for grammar, while
derivational morphology forms entirely new words. Both are essential for
understanding how language evolves and conveys meaning.

State differences between inflectional and

derivational morpheme
Here are the key differences between inflectional and derivational
morphemes:

Feature Inflectional Morpheme Derivational Morpheme

Expresses grammatical Forms new words or changes

Function
features word class

Meaning Does not change the core

Often changes the meaning
Change meaning

May change the word class (e.g.,

Word Class Keeps the same word class
noun → adjective)
Feature Inflectional Morpheme Derivational Morpheme

Usually added at the end of

Position Can be a prefix or suffix
a word

Number per Only one inflectional Multiple derivational morphemes

Word morpheme per word can occur

happy → unhappiness, write →

Examples walk → walked, dog → dogs
writer

In short, inflectional morphemes fine-tune grammar, while derivational

morphemes build new words.

What is meant by cognitive computing? What

principles guide our social behaviour?
Cognitive computing refers to technology that mimics human thought
processes to solve complex problems. It uses artificial intelligence (AI), machine
learning, natural language processing, and data mining to simulate human
reasoning and decision-making. Cognitive systems learn from data, recognize
patterns, and improve over time, enabling applications like virtual assistants,
medical diagnostics, and customer support.
Principles that guide our social behavior include:
1. Social Norms – Unwritten rules about how to behave in society.
2. Morality and Ethics – Understanding right from wrong and acting
accordingly.
3. Empathy – Recognizing and sharing the feelings of others.
4. Respect and Tolerance – Valuing others' rights, beliefs, and opinions.
5. Reciprocity – The give-and-take nature of social interactions.
6. Cultural Values – Shared beliefs and practices that influence behavior.
7. Laws and Rules – Formal regulations that shape acceptable conduct.
These principles ensure harmony, cooperation, and mutual understanding in
society.

Explain TEXT SUMMARIZATION

Text summarization is the process of automatically shortening a text document
while preserving its main ideas and meaning. It helps in quickly understanding
large volumes of information by generating a concise and coherent version of
the original content.
There are two main types of text summarization:
1. Extractive Summarization: Selects key sentences or phrases directly
from the original text and combines them to form a summary.
Example: From a news article, it might pull out important lines to
highlight the who, what, when, and where.
2. Abstractive Summarization: Generates new sentences that capture the
essence of the text, using natural language techniques.
Example: Instead of copying, it rewrites content in a more concise and
fluent way, like how a human might summarize.
Text summarization is widely used in news aggregation, search engines,
academic research, and customer service to make information easier and faster
to consume.

What is Sentiment Analysis? Types of Sentiment

Analysis
Sentiment Analysis is a Natural Language Processing (NLP) technique used to
identify and interpret emotions or opinions expressed in text. It classifies
content as positive, negative, or neutral, helping organizations understand
customer attitudes and feedback from sources like reviews, social media, and
surveys.
There are several types of sentiment analysis:
1. Binary Sentiment Analysis: Classifies text as either positive or negative.
2. Multiclass Sentiment Analysis: Includes a neutral category along with
positive and negative.
3. Fine-Grained Sentiment Analysis: Offers more detailed levels like very
positive, positive, neutral, negative, and very negative.
4. Aspect-Based Sentiment Analysis: Focuses on specific features within
the text (e.g., “The screen is great, but the battery is poor”).
5. Emotion Detection: Identifies specific emotions such as happiness,
anger, or sadness.
Sentiment analysis helps businesses gain insights into customer satisfaction,
improve products, and shape marketing strategies by understanding public
opinion at scale.
SUB : Information security

1. Explain RSA Algorithm

RSA (Rivest–Shamir–Adleman) is one of the first public-key cryptosystems and is widely used
for secure data transmission. It is based on the difficulty of factoring the product of two large
prime numbers.
Steps in RSA Algorithm:
Key Generation:
Select two distinct large prime numbers, p and q. Compute n = p × q. (n is used as the
modulus for both public and private keys.)

Compute Euler’s totient function: ϕ(n) = (p - 1) × (q -

1).

Choose an integer e such that 1 < e < ϕ(n) and gcd(e,

ϕ(n)) = 1; e is the public exponent.

Determine d as d ≡ e⁻¹ mod ϕ(n), meaning d is the modular multiplicative inverse of e mod
ϕ(n). d is the
private exponent.
Encryption:
The sender encrypts the message M using the receiver’s public key (e, n).
Ciphertext C = M^e mod n. Decryption:
The receiver decrypts the ciphertext C using their private key (d, n).
Message M = C^d mod n. Applications of RSA:
Secure data transmission Digital signatures
SSL/TLS protocols
2. What is Cipher Text? Explain its Types
Cipher Text is the encrypted form of a message. It appears scrambled and unreadable to
anyone who does not have the proper key to decrypt it.
Types of Cipher Text:
Block Cipher:
Encrypts fixed-size blocks of plaintext (e.g., 128 bits).
Each block is independently processed. Example: AES (Advanced Encryption Standard), DES
(Data Encryption Standard).
Stream Cipher:
Encrypts plaintextonebitoronebyteata time.
Suitable for real-time systems where data comes in streams.
Example: RC4, Salsa20. Key Difference:
Block ciphers work on blocks of data; stream ciphers work continuously on streams of data.
3. Explain Typesof Stream Cipher
Stream ciphers encrypt plaintext digits one at a time using a corresponding keystream digit.
Types:
Synchronous Stream Cipher:
The keystream is generated independently of the plaintext and ciphertext.
If synchronization between sender and receiver is lost, decryption will fail.
Example: RC4.
Advantage: Fast processing. Disadvantage: Sensitive to loss of
synchronization.
Self-Synchronizing Stream Cipher:
The keystream depends on a fixed number of previous ciphertext bits.
Even if some bits are lost, synchronization can be re-established.
Example: Cipher Feedback Mode (CFB). Advantage: Automatically resynchronizes after
errors.
4. Explain Steps in MD5 Algorithm
The MD5 (Message Digest 5) algorithm is used to produce a 128-bit hash value from any
length of data.
Steps:
Padding:
Add a ‘1’ bit to the message followed by ‘0’ bits. Make the length of the message congruent to
448 modulo 512 (i.e., 64 bits less than a multiple of 512). Appending Length:
Add a 64-bit representation of the original message length (before padding) to the result.
Initialize MD Buffer:
Initialize four 32-bit variables (A, B, C, D) with specific constants.
Processing Message in 512-bit Blocks:
Each 512-bit block is divided into 16 words of 32 bits.
Four rounds of operations using nonlinear functions and bitwise operations are performed on
each block. Output:
Thefinal result afterall blocks are processed isa concatenation of A, B, C, and D, giving a 128-
bit hash.
5. How Do You Create a Digital Signature
ADigital Signature isa mathematical technique usedto validate theauthenticity and
integrityof a message or document.
Steps:
Hashing the Message:
The original message is hashed using a cryptographic hash function (e.g., SHA-256).
Encrypting the Hash:
The hash value is encrypted using the sender's private key to create the digital signature.
Sending the Message and Signature: Both the original message and the
digital signature are sent to the receiver.
Verification by Receiver:
The receiver decrypts the signature using the sender’s public key to obtain the hash.
The receiver also hashes the received message separately.
If both hashes match, the message isauthentic and has not been tampered with.
Applications:
Secure email (like S/MIME) Electronic contracts
Blockchain transactions
6. Explain MAC (Message Authentication Code)
A Message Authentication Code (MAC) is a short piece of information used to authenticate
a message and to provide integrityand authenticity assurances.
How MAC Works:
The sender and receiver share a secret key. The sender inputs the message and the secret
key into a MAC algorithm to generate a MAC tag.
The receiver does the same process and checks if the calculated MAC matches the received
MAC.
If they match, the message is authentic.
Types of MACs:
HMAC (Hash-Based MAC): Uses a cryptographic hash function (like SHA-256) and a secret key.
CMAC(Cipher-Based MAC): Uses block cipher (like AES) algorithms.
Purpose:
Protects against tampering. Verifiesthesender's authenticity.
SUB: Image Processing Term Work

Q1. Write advantages and disadvantages of digital images

Ans: Advantages of Digital Images:

1. Easy Storage and Sharing

o Digital images can be stored in large numbers on small devices
and shared instantly via internet and cloud services.
2. Editing Flexibility
o Software tools allow easy editing, retouching, and enhancement
without degrading the original image.
3. High Quality and Resolution
o Modern digital cameras and smartphones can capture very high-
resolution images with great detail.
4. Cost-Effective
o Once you have a camera or device, taking and storing photos has
almost no additional cost compared to traditional film
photography.
5. Immediate Access
o Images can be viewed, edited, and printed immediately after
capture without waiting for film development.
6. Environmentally Friendly
o No need for chemical processing (used in film development),
reducing environmental harm.
7. Integration with Other Technologies
o Digital images can be used easily in multimedia applications,
websites, and AI systems (e.g., facial recognition).

Disadvantages of Digital Images:

1. Storage Limitations
o Very high-quality images consume a lot of storage space, leading
to the need for external drives or cloud subscriptions.
2. Risk of Loss or Corruption
o Digital files can be accidentally deleted, corrupted, or lost due to
hardware failure.
3. Quality Degradation with Compression
o Formats like JPEG use compression that can reduce image quality,
especially after multiple saves.
4. Privacy Concerns
o Easy sharing can lead to unauthorized distribution or misuse of
personal images.
5. Dependence on Technology
o Viewing or editing requires electronic devices and software, unlike
printed photos which can be seen without any tools.
6. Cost of Equipment
o High-end cameras, editing software, and storage solutions can be
expensive.
7. Over-Reliance on Editing
o Sometimes excessive digital manipulation can result in loss of
originality or authenticity of images.

Q2. Write a short note on 1. Hue 2. Mach band effect Ans:

1. Hue:
Hue is the attribute of a color that allows us to classify it as red, blue, green,
yellow, etc. It is determined by the wavelength of light emitted or reflected by
an object.
In color models:

 In the HSL (Hue, Saturation, Lightness) model, hue is represented as an

angle on a color wheel ranging from 0° to 360°.
o 0° corresponds to Red,
o 120° to Green,
o 240° to Blue, and so on.
 Hue does not account for how light or dark (brightness) or how vivid
(saturation) a color is — it only refers to the "basic color" identity.

For example:

 Pure red, pure green, and pure blue are hues.

 Pink is not a hue; it’s a tint (light version) of the hue red.

Applications of Hue:

 Image editing software allows users to adjust the hue to change colors in
photographs.
 Hue analysis is important in computer vision tasks like object detection
and tracking.

2. Mach Band Effect

The Mach Band Effect is a phenomenon observed in human visual perception

where a bright or dark band appears near the edges of regions with gradual
changes in brightness.

 It is named after Ernst Mach, an Austrian physicist who studied these

illusions.
 Although the brightness gradient in the image is smooth, the human eye
perceives sharper edges than actually exist.

Example:

 If you look at a grayscale gradient or a blurred shadow, you might notice

"extra" lines or bands that aren't actually present.

Importance of Mach Band Effect:

 It explains why our eyes are very sensitive to edges — crucial for
recognizing objects.
 In digital imaging, it can sometimes cause unwanted artifacts (false
bands) when smooth gradients are rendered.

Q3. In digital image processing, what is histogram processing?

Ans: - Histogram processing refers to techniques that use the histogram of an

image to improve its appearance or extract useful information.
A histogram in image processing is a graphical representation that shows the
frequency of occurrence of different pixel intensity values (brightness levels)
in an image.

 X-axis: Possible intensity values (e.g., 0 to 255 for an 8-bit grayscale

image).
 Y-axis: Number of pixels for each intensity value.
Main Purposes of Histogram Processing:

1. Image Enhancement
o Improve the contrast and visibility of features in an image.
2. Image Analysis
o Understand the brightness and contrast distribution.
o Detect underexposed (too dark) or overexposed (too bright)
images.

Common Histogram Processing Techniques:

1. Histogram Equalization
o Redistributes pixel intensities so that the histogram becomes
more uniform.
o Enhances the global contrast of an image, especially when the
usable data is concentrated in a narrow range.
2. Histogram Stretching (Contrast Stretching)
o Expands the range of intensity values to cover the full spectrum
(e.g., 0 to 255), improving contrast.
3. Histogram Specification (Matching)
o Adjusts an image so that its histogram matches a specified
histogram (useful in pattern recognition and medical imaging).

Example:

 A low-contrast image may have a histogram that is narrow and centered.

 After histogram equalization, the pixel values spread out over a wider
range, making the image appear clearer and sharper.

Q4. In homomorphic filtering, which high pass filter is used?

Ans: - In homomorphic filtering (used mainly for image enhancement like

illumination correction), a high-pass filter is applied in the frequency domain
to suppress low-frequency components (like uneven lighting) and enhance
high-frequency components (like edges and details).

The most commonly used high-pass filters in homomorphic filtering are:

1. Gaussian High-Pass Filter

 Formula:

H(u,v)=1−e−D(u,v)22D02H(u,v) = 1 - e^{-\frac{D(u,v)^2}{2D_0^2}}

 Where:
o D(u,v)D(u,v) is the distance from the origin of the frequency plane,
o D0D_0 is the cutoff frequency.
 Features:
o Smoothly removes low frequencies without sharp cut-offs.
o Avoids ringing artifacts (blurring around edges).

2. Butterworth High-Pass Filter

 Formula:

H(u,v)=11+(D0D(u,v))2nH(u,v) = \frac{1}{1 +
\left(\frac{D_0}{D(u,v)}\right)^{2n}}

 Where:
o nn is the order of the filter (controls the sharpness of transition),
o D0D_0 is the cutoff frequency.
 Features:
o Provides a smoother transition than an ideal filter,
o Order nn controls the steepness of the filter roll-off

Q5. Give the drawbacks of DFT.

Ans: Drawbacks of DFT:

1. High Computational Complexity

o The basic DFT requires O(N2)O(N^2)O(N2) operations for an input
of size NNN, making it slow for large datasets.
o Although the Fast Fourier Transform (FFT) improves this, the basic
DFT is computationally expensive.
2. Assumes Periodicity
o DFT treats the input signal or image as periodic, which can cause
discontinuities at the edges if the actual data is not periodic.
o These discontinuities introduce artifacts called spectral leakage.
3. Poor Localization in Time/Space
o DFT provides excellent frequency information but no time (or
spatial) information.
o It tells what frequencies are present but not where they occur in
the signal or image.
4. Aliasing Issues
o If the signal is not properly sampled (not satisfying the Nyquist
sampling rate), DFT can suffer from aliasing, causing distortion.
5. Not Suitable for Non-Stationary Signals
o DFT assumes that signal characteristics do not change over time
(stationarity).
o For signals whose properties vary over time (non-stationary
signals), DFT analysis can be misleading.
6. Memory Intensive
o For large images or signals, the storage of complex numbers for
each frequency component requires a lot of memory.
7. Edge Effects
o Especially in image processing, abrupt edges (boundaries) lead to
unwanted high-frequency components due to the DFT’s
assumption of continuity.

Q6. Handwritten and Printed Character Recognition

Ans: - Character Recognition refers to the process of automatically identifying

and converting images of letters, numbers, or symbols into machine-readable
text.
It is a part of a broader field called Optical Character Recognition (OCR).

Types:

1. Printed Character Recognition

o Focuses on recognizing typed or printed text from scanned
documents or images.
o Printed characters are usually uniform in size, shape, and spacing,
making recognition relatively easier.
2. Handwritten Character Recognition
o Focuses on recognizing handwritten text, either from paper
(offline) or from touchscreens (online).
o Handwriting is highly variable — different people write
differently, even the same person might write differently at
different times.
o This makes handwritten recognition more challenging.

Steps Involved in Character Recognition:

1. Preprocessing
o Cleaning the input image (noise removal, binarization, resizing).
o Aligning or normalizing characters.
2. Segmentation
o Separating individual characters or words from the background.
3. Feature Extraction
o Identifying important details such as strokes, curves, or edges that
represent the character shape.
4. Classification
o Matching the extracted features against known characters using
machine learning models (like SVMs, Neural Networks).
5. Post-processing
o Correcting minor errors using dictionaries, grammar rules, or
context.

Techniques Used:

 Template Matching
o Comparing the character image to stored patterns.
 Machine Learning Approaches
o Support Vector Machines (SVM)
o k-Nearest Neighbors (k-NN)
 Deep Learning Approaches
o Convolutional Neural Networks (CNNs) are very effective,
especially for complex handwritten text.
 Hidden Markov Models (HMMs)
o Used in recognizing cursive handwriting where characters are
connected.

Applications:
 Digitizing historical documents
 Automatic number plate recognition (vehicles)
 Bank check processing
 Handwritten note recognition on mobile devices
 Passport, ID card scanning

Challenges:

 Variation in handwriting styles

 Overlapping or joined characters (especially in cursive writing)
 Poor image quality (blurring, low resolution)
 Different fonts and layouts in printed text

Q7. Explain Morphological Operation

Ans: - Morphological operations are a set of image processing techniques that

process images based on their shapes.
They apply a structuring element (a small shape or kernel) to an image to
probe and transform the structure of objects in the image.

Morphological operations are mainly used on binary images (black and white),
but they can also be extended to grayscale images.

Basic Morphological Operations:

1. Erosion (⊖)
o Function: Shrinks or thins objects in the image.
o Effect:
 Removes small white noises.
 Breaks apart narrow connections between objects.
o How:
 The structuring element slides over the image and erodes
away boundaries of foreground (white) regions.
2. Dilation (⊕)
o Function: Expands or grows objects in the image.
o Effect:
 Fills small holes and connects nearby objects.
o How:
 The structuring element slides over the image and adds
pixels to the boundaries of foreground regions.
3. Opening (∘)
o Opening = Erosion followed by Dilation
o Function: Removes small objects or noise while keeping the
overall shape intact.
o Use: Smoothens the contour of objects and breaks narrow
bridges.
4. Closing (•)
o Closing = Dilation followed by Erosion
o Function: Fills small holes and gaps inside objects.
o Use: Useful for closing small breaks or gaps in the contours.

Structuring Element:

 A small pattern (like a 3×3 square, cross, or circle) used to probe the
image.
 The choice of structuring element size and shape affects the result of the
operation.

Applications of Morphological Operations:

 Noise removal (preprocessing)

 Object boundary extraction
 Shape analysis
 Skeletonization (thinning an object to a single-pixel wide structure)