Image Processing and Computer Vision (CS-803 (A)) - Unit 1 Short Notes

How unit 1 is examined

This unit covers what CVIP is, its models, filtering, digitisation and the whole of binary and gray-scale morphology; dilation and thinning carry the most marks, then the medium topics (basics, CV models, filtering, representations).

Basics of CVIP

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Medium weight</span>

Definition. <mark>Computer vision is the science of making a machine understand a scene from images, while image processing takes an image in and gives an improved or transformed image out.</mark>

Key points.

1. Image processing is a low-level activity: input and output are both images, as in noise removal, enhancement and restoration.
2. Computer vision is a high-level activity: the input is an image and the output is a description or decision, such as "this is a face" or "the obstacle is 2 m away".
3. Image processing is needed by vision as pre-processing, because a noisy, blurred or badly lit image gives wrong segmentation and recognition.
4. The usual chain is acquisition, pre-processing, segmentation, feature extraction, recognition and interpretation.
5. Applications include medical imaging, face and fingerprint recognition, remote sensing, robotics and autonomous vehicles.

Diagram. A digital image processing system has five elements.

<figure class="ds-fig" style="margin:1.4rem 0;overflow-x:auto"><svg xmlns="http://www.w3.org/2000/svg" id="dsfig-u1-01" viewBox="0 0 596 80" width="596" height="80" role="img" aria-label="Elements of a digital image processing system: Acq acquisition (sensor and digitiser), Sto storage, Pro processing, Com communication, Dis display"><style>#dsfig-u1-01 .e{stroke:#454C5A;stroke-width:1.4;fill:none}#dsfig-u1-01 .e.hi{stroke:#2340B8;stroke-width:2.6}#dsfig-u1-01 .n{fill:#FFFFFF;stroke:#16181D;stroke-width:1.4}#dsfig-u1-01 .n.hi{fill:#E3E9FC;stroke:#2340B8;stroke-width:2.2}#dsfig-u1-01 .n.rb-b{fill:#16181D;stroke:#16181D}#dsfig-u1-01 .n.rb-r{fill:#BD3227;stroke:#BD3227}#dsfig-u1-01 text{font-family:"JetBrains Mono",ui-monospace,Menlo,Consolas,monospace;font-size:13px}#dsfig-u1-01 .t{fill:#16181D;font-weight:500}#dsfig-u1-01 .t.inv{fill:#FFFFFF;font-weight:700}#dsfig-u1-01 .kd{stroke:#16181D;stroke-width:1.2}#dsfig-u1-01 .dot{fill:#16181D}#dsfig-u1-01 .ann{fill:#2340B8;font-size:11px;font-weight:700}#dsfig-u1-01 .lbl{fill:#6F7787;font-family:system-ui,-apple-system,sans-serif;font-size:12px;font-weight:700}#dsfig-u1-01 .ptr{fill:#2340B8;font-size:12px;font-weight:700}#dsfig-u1-01 .ah{fill:#454C5A}#dsfig-u1-01 .ah.hi{fill:#2340B8}#dsfig-u1-01 .wl rect{fill:#FFFFFF;stroke:#DCE0E7}#dsfig-u1-01 .wl .t{font-size:12px;font-weight:700}#dsfig-u1-01 .wl.hi rect{fill:#2340B8;stroke:#2340B8}#dsfig-u1-01 .wl.hi .t{fill:#FFFFFF}html.dark #dsfig-u1-01 .e{stroke:#B1B7C3}html.dark #dsfig-u1-01 .e.hi{stroke:#8FA3FF}html.dark #dsfig-u1-01 .n{fill:#161920;stroke:#E6E8ED}html.dark #dsfig-u1-01 .n.hi{fill:#1E2748;stroke:#8FA3FF}html.dark #dsfig-u1-01 .n.rb-b{fill:#E6E8ED;stroke:#E6E8ED}html.dark #dsfig-u1-01 .n.rb-r{fill:#FF7E71;stroke:#FF7E71}html.dark #dsfig-u1-01 .t{fill:#E6E8ED}html.dark #dsfig-u1-01 .t.inv{fill:#0F1115}html.dark #dsfig-u1-01 .kd{stroke:#E6E8ED}html.dark #dsfig-u1-01 .dot{fill:#E6E8ED}html.dark #dsfig-u1-01 .ann{fill:#8FA3FF}html.dark #dsfig-u1-01 .lbl{fill:#858D9C}html.dark #dsfig-u1-01 .ptr{fill:#8FA3FF}html.dark #dsfig-u1-01 .ah{fill:#B1B7C3}html.dark #dsfig-u1-01 .ah.hi{fill:#8FA3FF}html.dark #dsfig-u1-01 .wl rect{fill:#161920;stroke:#2A2E37}html.dark #dsfig-u1-01 .wl.hi rect{fill:#8FA3FF;stroke:#8FA3FF}html.dark #dsfig-u1-01 .wl.hi .t{fill:#0F1115}</style><defs><marker id="ah1" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah" d="M0,1 L9,5 L0,9 z"/></marker><marker id="ahh1" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah hi" d="M0,1 L9,5 L0,9 z"/></marker></defs><path class="e" d="M59,40 L148,40" marker-end="url(#ah1)"/><path class="e" d="M188,40 L277,40" marker-end="url(#ah1)"/><path class="e" d="M317,40 L406,40" marker-end="url(#ah1)"/><path class="e" d="M446,40 L535,40" marker-end="url(#ah1)"/><circle class="n" cx="40" cy="40" r="18"/><text class="t" x="40" y="40" dy=".35em" text-anchor="middle">Acq</text><circle class="n" cx="169" cy="40" r="18"/><text class="t" x="169" y="40" dy=".35em" text-anchor="middle">Sto</text><circle class="n" cx="298" cy="40" r="18"/><text class="t" x="298" y="40" dy=".35em" text-anchor="middle">Pro</text><circle class="n" cx="427" cy="40" r="18"/><text class="t" x="427" y="40" dy=".35em" text-anchor="middle">Com</text><circle class="n" cx="556" cy="40" r="18"/><text class="t" x="556" y="40" dy=".35em" text-anchor="middle">Dis</text></svg><figcaption style="font-size:.82em;opacity:.72;margin-top:.45rem">Elements of a digital image processing system: Acq acquisition (sensor and digitiser), Sto storage, Pro processing, Com communication, Dis display</figcaption></figure>

Elements.

1. Acquisition uses a sensor (camera, scanner) and a digitiser to convert the scene into a digital image.
2. Storage holds the image in short-term memory (frame buffer), online disks and archival media.
3. Processing is the computer with software that enhances, restores, segments or compresses the image.
4. Communication sends images over networks, so compression is used to save bandwidth.
5. Display shows the result on monitors or printers.

Answer frame. Open with the definition of computer vision and how it differs from image processing; draw the five-block diagram; develop points 1-5 (low versus high level, then why processing is required); close with "image processing prepares the data and vision interprets it". For the system-elements question, draw the block diagram first and give one sentence per element.

Asked: [7 marks] (May 2022, May 2024) Describe the concept of computer vision which is mandatory in image processing. Asked: [7 marks] (May 2022) Describe various elements of digital image processing system. Asked: [7 marks] (May 2024) What are the components of digital image processing system? Explain each in detail.

History of CVIP

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. History of CVIP is the timeline of how image processing (improving images) and computer vision (understanding images) developed and merged.

Key points.

1. In the 1920s-50s pictures were sent by cable and early scanners appeared; the 1960s space programme (Ranger and Apollo images) made computer-based enhancement important.
2. In 1963 Roberts' thesis on recognising 3-D blocks from 2-D images started computer vision.
3. In the 1970s Marr's theory of vision, edge detectors and medical CT scanning appeared.
4. The 1980s brought active vision, stereo, motion analysis and industrial inspection, and the 1990s brought statistical methods, face recognition and cheap digital cameras.
5. Since 2012 deep learning (CNNs such as AlexNet) has dominated both fields.

Asked: [7 marks] (Dec 2024) Discuss the history and evolution of Computer Vision and Image Processing (CVIP).

Evolution of CVIP

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. Evolution of CVIP is the shift from hand-crafted image operations to learned, intelligent scene understanding, driven by advances in hardware, sensors and algorithms.

Key points.

1. The stages are early digital image processing, feature-based vision, statistical and machine learning methods, and deep learning.
2. Hardware advances (faster CPUs, GPUs, large memory) made real-time processing of large images possible.
3. Better sensors (CCD and CMOS, depth cameras, satellites, medical scanners) supplied high-resolution and 3-D data.
4. Algorithms moved from filters and edges to SIFT and SVM, and then to CNNs that learn features automatically.
5. The impact is that applications grew from enhancement to self-driving cars, medical diagnosis and face recognition, and the boundary between CV and IP has blurred.

Asked: [7 marks] (Jun 2025) Discuss the evolution and history of CVIP. How have advancements in technology influenced the development of CVIP?

CV Models

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Medium weight</span>

Definition. <mark>A computer vision model is a mathematical description of how a scene forms an image (geometry, light, colour) and how that image can be interpreted.</mark>

Key points.

1. Geometric models describe image formation: the pinhole camera maps a 3-D point $(X,Y,Z)$ to $x=fX/Z,\ y=fY/Z$ and gives perspective and reconstruction.
2. Photometric models describe how illumination, surface reflectance and viewing angle decide pixel brightness, and are used for shape from shading.
3. Shape (object) models represent objects as 2-D or 3-D structures such as wireframes, generalised cylinders and CAD models, used in recognition.
4. Statistical and learning-based models (Bayes classifiers, SVM, CNN) learn the appearance of classes from data.
5. Applications are recognition, 3-D reconstruction, robot navigation and inspection.

Colour vision models.

Colour models are needed because RGB mixes intensity with colour, whereas HSI separates them and matches human perception. Conversion is $I=(R+G+B)/3$.

Answer frame. Open with the definition; list geometric, photometric, shape and statistical models with one line each and an application; for the colour question start with the need for colour models, draw the table above and end by comparing RGB with HSI; close with the applications.

Asked: [7 marks] (May 2022, May 2023) Describe various computer vision models and their applications / different types of computer vision models. Asked: [7 marks] (May 2024) Discuss various color vision models.

Image Filtering

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Medium weight</span>

Definition. <mark>Image filtering modifies pixel values using a neighbourhood or frequency operation to remove noise or enhance features.</mark>

Key points.

1. Spatial filtering works directly on pixels: $g(x,y)=\sum_s\sum_t w(s,t)f(x+s,y+t)$ with a mask $w$.
2. Smoothing filters (mean, Gaussian, median) reduce noise; the median filter removes salt-and-pepper noise best and preserves edges.
3. Sharpening filters (Laplacian, unsharp masking, Sobel gradient) highlight edges and fine detail.
4. Frequency filtering follows $G(u,v)=H(u,v)F(u,v)$: take the DFT, multiply by the filter, take the inverse DFT.
5. Low-pass filters (ideal, Butterworth, Gaussian) smooth; ideal causes ringing, Gaussian has none, and Butterworth is in between with order $n$.
6. High-pass filters sharpen, and homomorphic filtering treats $f=i\cdot r$ by taking the log, then filtering to reduce illumination and boost reflectance.

Answer frame. Open with the definition and purpose; draw the DFT, filter, inverse DFT block chain for the frequency question; develop spatial then frequency techniques with examples; close with the table. For the "distinguish" question write the table first.

Asked: [7 marks] (May 2023) Discuss in detail the various types of image filtering techniques. Asked: [7 marks] (May 2024) Show the various techniques in frequency domain to enhance an image with necessary examples. Asked: [7 marks] (May 2024) Distinguish between spatial and frequency domain image enhancement.

Image Representations

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Medium weight</span>

Definition. <mark>A digital image is a two-dimensional function $f(x,y)$ whose spatial coordinates and amplitude are both discrete, stored as an $M\times N$ matrix.</mark>

Key points.

1. A continuous image is converted by sampling followed by quantization.
2. Sampling discretises the spatial coordinates $x,y$ into $M\times N$ pixels; more samples give higher spatial resolution.
3. Quantization discretises the intensity into $L=2^k$ gray levels; more bits $k$ give more gray levels and less false contouring.
4. Coarse sampling causes checkerboard patterns and coarse quantization causes false contours.
5. Storage size is $b=M\times N\times k$ bits, so a $256\times256$ 8-bit image needs $524288$ bits, which is $64$ KB.
6. An image is represented as a matrix, or as binary, gray-scale, colour or indexed types, and transforms (Fourier, DCT, wavelet) give another representation.
7. Image transformation converts an image into another domain to give decorrelation, energy compaction, easier filtering and compression.

<figure class="ds-fig" style="margin:1.4rem 0;overflow-x:auto"><svg xmlns="http://www.w3.org/2000/svg" id="dsfig-u1-02" viewBox="0 0 596 80" width="596" height="80" role="img" aria-label="Digitisation: scene, sensor (continuous image), sampling (spatial), quantization (gray levels), digital image matrix"><style>#dsfig-u1-02 .e{stroke:#454C5A;stroke-width:1.4;fill:none}#dsfig-u1-02 .e.hi{stroke:#2340B8;stroke-width:2.6}#dsfig-u1-02 .n{fill:#FFFFFF;stroke:#16181D;stroke-width:1.4}#dsfig-u1-02 .n.hi{fill:#E3E9FC;stroke:#2340B8;stroke-width:2.2}#dsfig-u1-02 .n.rb-b{fill:#16181D;stroke:#16181D}#dsfig-u1-02 .n.rb-r{fill:#BD3227;stroke:#BD3227}#dsfig-u1-02 text{font-family:"JetBrains Mono",ui-monospace,Menlo,Consolas,monospace;font-size:13px}#dsfig-u1-02 .t{fill:#16181D;font-weight:500}#dsfig-u1-02 .t.inv{fill:#FFFFFF;font-weight:700}#dsfig-u1-02 .kd{stroke:#16181D;stroke-width:1.2}#dsfig-u1-02 .dot{fill:#16181D}#dsfig-u1-02 .ann{fill:#2340B8;font-size:11px;font-weight:700}#dsfig-u1-02 .lbl{fill:#6F7787;font-family:system-ui,-apple-system,sans-serif;font-size:12px;font-weight:700}#dsfig-u1-02 .ptr{fill:#2340B8;font-size:12px;font-weight:700}#dsfig-u1-02 .ah{fill:#454C5A}#dsfig-u1-02 .ah.hi{fill:#2340B8}#dsfig-u1-02 .wl rect{fill:#FFFFFF;stroke:#DCE0E7}#dsfig-u1-02 .wl .t{font-size:12px;font-weight:700}#dsfig-u1-02 .wl.hi rect{fill:#2340B8;stroke:#2340B8}#dsfig-u1-02 .wl.hi .t{fill:#FFFFFF}html.dark #dsfig-u1-02 .e{stroke:#B1B7C3}html.dark #dsfig-u1-02 .e.hi{stroke:#8FA3FF}html.dark #dsfig-u1-02 .n{fill:#161920;stroke:#E6E8ED}html.dark #dsfig-u1-02 .n.hi{fill:#1E2748;stroke:#8FA3FF}html.dark #dsfig-u1-02 .n.rb-b{fill:#E6E8ED;stroke:#E6E8ED}html.dark #dsfig-u1-02 .n.rb-r{fill:#FF7E71;stroke:#FF7E71}html.dark #dsfig-u1-02 .t{fill:#E6E8ED}html.dark #dsfig-u1-02 .t.inv{fill:#0F1115}html.dark #dsfig-u1-02 .kd{stroke:#E6E8ED}html.dark #dsfig-u1-02 .dot{fill:#E6E8ED}html.dark #dsfig-u1-02 .ann{fill:#8FA3FF}html.dark #dsfig-u1-02 .lbl{fill:#858D9C}html.dark #dsfig-u1-02 .ptr{fill:#8FA3FF}html.dark #dsfig-u1-02 .ah{fill:#B1B7C3}html.dark #dsfig-u1-02 .ah.hi{fill:#8FA3FF}html.dark #dsfig-u1-02 .wl rect{fill:#161920;stroke:#2A2E37}html.dark #dsfig-u1-02 .wl.hi rect{fill:#8FA3FF;stroke:#8FA3FF}html.dark #dsfig-u1-02 .wl.hi .t{fill:#0F1115}</style><defs><marker id="ah2" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah" d="M0,1 L9,5 L0,9 z"/></marker><marker id="ahh2" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah hi" d="M0,1 L9,5 L0,9 z"/></marker></defs><path class="e" d="M59,40 L148,40" marker-end="url(#ah2)"/><path class="e" d="M188,40 L277,40" marker-end="url(#ah2)"/><path class="e" d="M317,40 L406,40" marker-end="url(#ah2)"/><path class="e" d="M446,40 L535,40" marker-end="url(#ah2)"/><circle class="n" cx="40" cy="40" r="18"/><text class="t" x="40" y="40" dy=".35em" text-anchor="middle">Sc</text><circle class="n" cx="169" cy="40" r="18"/><text class="t" x="169" y="40" dy=".35em" text-anchor="middle">Sen</text><circle class="n" cx="298" cy="40" r="18"/><text class="t" x="298" y="40" dy=".35em" text-anchor="middle">Smp</text><circle class="n" cx="427" cy="40" r="18"/><text class="t" x="427" y="40" dy=".35em" text-anchor="middle">Qnt</text><circle class="n" cx="556" cy="40" r="18"/><text class="t" x="556" y="40" dy=".35em" text-anchor="middle">Img</text></svg><figcaption style="font-size:.82em;opacity:.72;margin-top:.45rem">Digitisation: scene, sensor (continuous image), sampling (spatial), quantization (gray levels), digital image matrix</figcaption></figure>

Answer frame. Open with the definition of a digital image; draw the digitisation chain and a small matrix; explain sampling then quantization with resolution effects; close with the storage formula. For the significance question, say representation decides what filtering can do and add filtering (noise removal, enhancement).

Asked: [7 marks] (May 2022, May 2024) Define digital image. What do you mean by image sampling and quantization? Illustrate how the image is digitized. Asked: [7 marks] (Dec 2024) Explain the significance of image representations and image filtering. Asked: [7 marks] (May 2023) What is meant by Image Transformation? Explain its needs in digital image processing.

Image Statistics

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Image statistics are numerical measures of the gray-level distribution, taken from the histogram $p(r_k)=n_k/n$.

Key points.

1. Mean gray level $\mu=\sum r_kp(r_k)$ shows overall brightness.
2. Variance $\sigma^2=\sum (r_k-\mu)^2p(r_k)$ shows contrast.
3. Skewness and entropy describe histogram shape and information content.
4. They guide thresholding, equalisation and noise estimation.

Conditioning

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. Conditioning is the pre-processing that cleans an image so that later stages work reliably.

Key points.

1. Conditioning steps are enhancement (contrast stretching, equalisation), filtering (noise removal) and normalisation (size, brightness, orientation).
2. Labeling then segments the image and assigns each region a label, typically with connected component labeling.
3. Grouping combines related pixels, edges or regions into objects.
4. The recognition pipeline is conditioning, labeling, grouping, extracting and matching.

Asked: [7 marks] (Jun 2025) Explain the steps involved in image conditioning, labeling, and grouping for image recognition.

Labeling

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Labeling assigns each pixel or region an identity, such as an object number.

Key points.

1. Pixels are first classified by thresholding or edge detection.
2. Connected pixels are given the same label using 4- or 8-connectivity.
3. Labels let each region be counted and measured separately.

Grouping

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Grouping gathers labeled pixels, edge segments or regions that belong together into one object.

Key points.

1. Similarity of intensity, colour, texture or proximity decides the grouping.
2. Examples are joining edge fragments into lines and merging regions.
3. The output is a set of objects ready for measurement.

Extracting

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Extracting measures properties (features) of each grouped object.

Key points.

1. Typical features are area, perimeter, centroid, moments and shape descriptors.
2. Good features are compact and invariant to shift, rotation and scale.
3. They form the feature vector for matching.

Matching

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Matching compares the extracted features with stored model features to recognise the object.

Key points.

1. Methods include template matching, minimum distance classification and structural matching.
2. The best-scoring model gives the object's identity.
3. A threshold rejects unknown objects.

Morphological Image Processing: Introduction

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. <mark>Morphological image processing is a set-theory based technique that probes an image with a small shape called a structuring element (SE) to extract or modify shapes.</mark>

Key points.

1. Binary images are treated as sets of foreground pixels, and gray-scale images use min and max.
2. The SE (for example a $3\times3$ square or cross) has an origin and moves over the image like a mask.
3. The primary operations are dilation, erosion, opening and closing; hit-or-miss, thinning and thickening are built from them.
4. Applications are noise removal, boundary extraction, skeletons, hole filling and shape analysis.

Asked: [7 marks] (Dec 2024) Define morphological image processing and describe its primary operations.

Dilation

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. ==Dilation of $A$ by structuring element $B$ is $A\oplus B=\{z\mid (\hat B)_z\cap A\neq\emptyset\}$: it grows objects.==

Key points.

1. The SE is reflected and slid over the image, and the output is 1 wherever it overlaps at least one foreground pixel.
2. It expands object boundaries, so objects get larger.
3. It fills small holes and gaps and joins nearby broken parts.
4. In gray-scale it is a local maximum: $(f\oplus b)(x,y)=\max\{f(x-s,y-t)+b(s,t)\}$, so bright regions grow and dark details shrink.
5. It is commutative and increasing, and it is the dual of erosion of the complement.
6. Bridging broken characters is a typical use.

Example. A $5\times5$ image has a $3\times3$ block of ones in the centre and the SE is a $3\times3$ cross.

Dilation versus erosion.

Answer frame. Open with the formula and meaning; give the 5x5 example as a small grid; develop points 1-4; for the four-operation question define dilation, erosion, closing and thinning with expressions in one or two lines each. For the differentiate question write the table and add binary and gray-scale rows.

Asked: [14 marks] (May 2022) Explain the following operations: i) dilation ii) erosion iii) closing iv) thinning. Asked: [7 marks] (Dec 2024) Differentiate between dilation and erosion in binary and grayscale images.

Erosion

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Erosion of $A$ by $B$ is $A\ominus B=\{z\mid B_z\subseteq A\}$: it shrinks objects.

Key points.

1. A pixel stays 1 only if the SE placed there fits entirely inside the object.
2. It removes thin lines and small noise specks and separates touching objects.
3. In gray-scale it is a local minimum, so bright details shrink.
4. Boundary extraction is $\beta(A)=A-(A\ominus B)$.

Opening

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Opening is erosion followed by dilation: $A\circ B=(A\ominus B)\oplus B$.

Key points.

1. It removes small objects, thin protrusions and noise while keeping the main shape size.
2. It smooths contours and breaks narrow connections.
3. It is idempotent: applying it twice equals applying it once.

Closing

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Closing is dilation followed by erosion: $A\bullet B=(A\oplus B)\ominus B$.

Key points.

1. It fills small holes and narrow gaps and joins near parts.
2. It smooths the contour from outside without much change in size.
3. It is idempotent and is the dual of opening.

Hit-or-Miss transformation

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Hit-or-miss detects a shape: $A\circledast B=(A\ominus B_1)\cap(A^c\ominus B_2)$, where $B_1$ is the foreground pattern and $B_2$ the background pattern.

Key points.

1. It outputs 1 only where the foreground pattern hits the object and the background pattern hits the background.
2. It finds specific shapes, corners and isolated points.
3. Thinning and thickening are built from it.

Morphological algorithm operations on binary images

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. Binary morphological algorithms apply set operations on foreground pixels using an SE.

Key points.

1. Dilation adds pixels around objects wherever the SE touches them, and erosion keeps only pixels where the SE fits.
2. Opening and closing combine them for noise removal and gap filling.
3. Further algorithms are boundary extraction, region filling, connected component extraction, convex hull, thinning, thickening and skeletons.
4. Dilation and erosion are duals: $(A\ominus B)^c=A^c\oplus\hat B$.

Asked: [7 marks] (Jun 2025) Define morphological image processing and describe the operations of dilation and erosion on binary images.

Morphological algorithm operations on gray-scale images

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. Gray-scale morphology uses a structuring function, with max replacing union and min replacing intersection.

Key points.

1. Dilation takes the local maximum of $f+b$ and erosion takes the local minimum of $f-b$.
2. Opening removes bright small details and closing removes dark small details.
3. Smoothing is opening followed by closing, and the morphological gradient is dilation minus erosion.
4. Top-hat is $f-(f\circ b)$ and shows bright objects on uneven background.
5. Reconstruction rebuilds shapes from a marker image under a mask.

Asked: [7 marks] (May 2022) Discuss morphological algorithm operations performed on gray scale images.

Thinning

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. ==Thinning removes boundary pixels of a binary object, without breaking connectivity, until a one-pixel-wide skeleton remains: $A\otimes B=A-(A\circledast B)$.==

Key points.

1. It uses the hit-or-miss transform to find boundary pixels that match a pattern, then subtracts them from the set.
2. A sequence of structuring elements $\{B\}=\{B^1,B^2,\dots,B^8\}$ (rotations of the pattern) is applied one after another: $A\otimes\{B\}=((A\otimes B^1)\otimes B^2)\dots\otimes B^n$.
3. The process repeats until no further change occurs.
4. It preserves the topology and the end points, so the result is the medial line of the shape.
5. Thickening is the dual: $A\odot B=A\cup(A\circledast B)$, which adds pixels to grow the object, and equals thinning of the background.
6. Applications are fingerprint ridge analysis, OCR stroke thinning, skeletonisation and road-map extraction.

Step 1: Pick the first rotation of the pattern B.
Step 2: Find pixels where A hit-or-miss B is true.
Step 3: Delete them from A.
Step 4: Repeat with the next rotation.
Step 5: Cycle through all rotations until A stops changing.

Answer frame. Open with the thinning definition and expression; write the algo steps; explain points 1-4 in order; for thinning and thickening add point 5 and an application such as fingerprints; close with "thinning gives skeletons for shape analysis". For the two-of-four question write the definition, steps and applications.

Asked: [14 marks] (May 2023) Explain in brief any two of the following: a) Thinning b) Motion-based segmentation c) View class matching d) Knowledge representation. Asked: [7 marks] (Jun 2025) Describe thinning and thickening operations. Provide an example of an application where these operations are useful.

Thickening

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Thickening is $A\odot B=A\cup(A\circledast B)$, the dual of thinning.

Key points.

1. It adds pixels to the object using hit-or-miss.
2. It is done with rotating structuring elements until stable.
3. It fills concavities and grows regions without merging them.

Region growing, region shrinking

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Region growing starts from seed pixels and adds neighbours that satisfy a similarity rule; region shrinking removes boundary pixels of a region.

Key points.

1. Region growing needs seeds, a similarity criterion (intensity difference below a threshold) and a stopping rule.
2. Growing is an iterative dilation restricted to similar pixels.
3. Shrinking is repeated erosion and reduces regions to points or skeletons.
4. Growing gives connected regions but depends on seed choice.

Last-minute revision

• Computer vision = understanding a scene; image processing = image in, image out.
• DIP system elements: acquisition, storage, processing, communication, display.
• Digital image: $f(x,y)$ sampled in space and quantized in amplitude; $L=2^k$; size $=M\times N\times k$.
• Frequency filtering: $G=HF$; ideal has ringing, Butterworth is the compromise, Gaussian has none.
• Dilation $A\oplus B$ grows (max); erosion $A\ominus B$ shrinks (min).
• Opening = erosion then dilation; closing = dilation then erosion.
• Hit-or-miss $=(A\ominus B_1)\cap(A^c\ominus B_2)$.
• Thinning $A-(A\circledast B)$; thickening $A\cup(A\circledast B)$.
• 3x3 block with cross SE: dilation gives 13 pixels, erosion gives 1.
• Gradient = dilation minus erosion; top-hat = $f-$ opening.
• Recognition chain: conditioning, labeling, grouping, extracting, matching.

Memory hooks

• "Dilate = Do more, Erode = Eat away."
• "Opening opens gaps (removes specks); closing closes holes."
• "Ideal rings, Butterworth bends, Gaussian glides."
• Pipeline mnemonic: "Clean, Label, Group, Extract, Match" (C-L-G-E-M).
• Sampling is space (x,y); quantization is amplitude (gray).

Coverage checklist

• Basics of CVIP: computer vision and image processing, DIP system elements (Q3, Q4, Q21)
• History of CVIP: timeline (Q10)
• Evolution of CVIP: stages and technology impact (Q9)
• CV Models: geometric, photometric, shape, colour models (Q5, Q6)
• Image Filtering: spatial, frequency, comparison (Q11, Q12, Q13)
• Image Representations: digital image, sampling, quantization, transformation (Q14, Q15, Q20)
• Image Statistics: mean, variance, histogram
• Conditioning: conditioning steps (Q7)
• Labeling: connected labeling (Q7)
• Grouping: grouping (Q7)
• Extracting: features
• Matching: recognition
• Morphological Image Processing: Introduction (Q16)
• Dilation (Q1, Q8)
• Erosion (Q1, Q8)
• Opening
• Closing (Q1)
• Hit-or-Miss transformation
• Morphological algorithm operations on binary images (Q17)
• Morphological algorithm operations on gray-scale images (Q18)
• Thinning (Q1, Q2, Q19)
• Thickening (Q19)
• Region growing, region shrinking