1) Process of 2D and 3D Geometric Transformations

2D Geometric Transformations
2D transformations change the position, size, or shape of objects on a flat image (x, y plane).
Types of 2D transformations:
1. Translation
o Moves an object from one place to another.
o Shape and size do not change.
o Example: Shifting an image left, right, up, or down.
2. Rotation + Translation (Rigid/Euclidean)
o Rotates the object and then moves it.
o Distances and angles remain the same.
o Example: Turning a photo slightly and placing it elsewhere.
3. Scaling + Rotation (Similarity)
o Changes the size of the object (bigger or smaller) and may rotate it.
o Shape remains the same.
o Example: Zooming in/out of an image.
4. Affine Transformation
o Includes scaling, rotation, translation, and shearing.
o Parallel lines remain parallel.
o Shape may change but straight lines stay straight.
5. Projective Transformation (Perspective/Homography)
o It makes close objects look bigger and far objects look smaller
o Straight lines remain straight, but parallel lines may meet.
o Used in camera images and 3D-to-2D projection.
These transformations are applied using matrices on coordinate points.
3D Geometric Transformations
3D transformations change objects in three-dimensional space (x, y, z).
Types of 3D transformations:
1. Translation
o Moves a 3D object to a new position.
o Example: Moving a cube forward or sideways.
2. Rotation + Translation (Rigid Motion)
o Rotates the object in space and then shifts it.
o Shape and size remain unchanged.
3. Scaling + Rotation (Similarity)
o Changes the size of the 3D object while preserving shape.
4. Affine Transformation
o Allows stretching, shrinking, or skewing in 3D.
o Parallel lines and planes remain parallel.
5. Projective Transformation
o Makes the image look real, the way a camera sees the world.
o Changes a 3D scene into a flat 2D picture.
These transformations help convert real-world 3D scenes into 2D images captured by
cameras.

2. Geometric Primitives and Their Role in Image Formation

Geometric primitives are the basic building blocks used to describe shapes in images and
scenes.
Main primitives:
1. Points
o Represent a location in space.
o In images, a point corresponds to a pixel.
2. Lines
o Formed by connecting two points.
o Used to represent edges and boundaries.
3. Planes
o Flat surfaces extending in space.
o Used to model walls, floors, or surfaces of objects.
4. Curves, Surfaces, and Volumes
o Used for more complex shapes.
Role in Image Formation
1. Describe the Real Scene
 They help represent the shapes and positions of objects in 3D space.
2. Help the Camera Understand Shapes
 Cameras capture light from these shapes to form an image.
3. Convert 3D to 2D
 Real-world objects are 3D, but images are flat.
 Primitives help project 3D shapes onto a 2D image.
4. Help Detect Edges and Corners
 Computers use points and lines to find object boundaries.
5. Used in Measurement
 Help calculate distance, size, and position of objects.
6. Important for Computer Vision Tasks
 Object detection
 Face recognition
 Self-driving cars
 3D modeling
7. Help Rebuild 3D Scenes
 From multiple images, computers can reconstruct the original 3D scene.
3. Lens Distrotation
 Most camera models assume that straight lines in the real world appear straight in the
image.
 But many lenses, especially wide-angle lenses, bend straight lines and make them
look curved. This effect is called lens distortion.
 Because of distortion, the image may look bent or not normal, and it becomes hard to
measure things correctly
 If distortion is not corrected, the images can look blurry because matching parts do
not align properly.
 The most common type of distortion is radial distortion, where image points move
inward or outward from the center of the image.
 There are two main kinds of radial distortion
1. Barrel distortion: Lines curve outward like a barrel (image bulges).
2. Pincushion distortion: Lines curve inward toward the center.
 This distortion depends on how far a point is from the center of the image. The farther
it is, the more it gets distorted.
 Special formulas (models) can be used to correct this distortion and make lines
straight again.
 Some lenses produce very strong distortion and show a very wide view (almost 180°).
They need special correction models.
 Very complex lenses may need more advanced methods to fix distortions accurately.
 After correcting distortion, the camera image behaves more like a normal linear image
where straight lines appear straight.

4. Sampling and Aliasing

 Sampling is the process of taking small pieces of the real image and recording them as
pixels in a digital image.
 The camera sensor has many tiny cells, and each cell measures the light falling on it.
 These measurements are converted into numbers to form the digital picture.
 If the sensor captures enough samples, the image looks clear and accurate.
 If it captures too few samples, important details are missed.
 Aliasing happens when fine patterns or details are recorded incorrectly because of low
sampling.
 It can cause zigzag edges, wavy lines, false colors, or repeated patterns in the image.
 For example, stripes on clothes may look distorted or flicker in photos or videos.
 To avoid aliasing, the sampling rate should be high (more pixels, better sensors).
 Blurring the image slightly before sampling can also reduce aliasing by removing
very fine details.
 Cameras sometimes use special anti-aliasing filters to smooth the image before it
reaches the sensor.
 Lens blur and sensor size also affect how much aliasing occurs.
  Aliasing is not only during image capture; it can also happen when resizing or
compressing images.
 Proper filtering methods are used in image processing to reduce these effects.

5. Point Spread Function (PSF)

1. Definition
1. The Point Spread Function (PSF) describes how an imaging system (camera) records
a single point of light.
2. Ideally, a point of light should appear as a sharp point in the image.
3. In real cameras, it appears as a small blurred spot due to lens and sensor effects.
4. PSF represents the shape and size of this blur.
2. Causes of PSF
1. Problems in the camera lens make the light spread out.
2. Each pixel collects light from a small area, not just one exact point.
3. If the camera is not focused properly, the image becomes more blurry.
4. Movement and surroundings can also make the light spread and blur the image.
3. What PSF Depends On
1. Lens quality and focus.
2. Pixel size, spacing, and fill factor of the sensor.
3. Presence of anti-aliasing filters.
4. Camera settings and distance from the object.
PSF is mainly the combined effect of lens blur and sensor properties.
4. Role of PSF in Images
1. Determines how sharp or blurry an image looks.
2. Affects how much detail can be seen.
3. Helps predict aliasing (image errors due to poor sampling).
4. Important for improving image quality.
5. Used in image restoration and deblurring methods.
5. PSF and Aliasing
1. A large blur (large PSF) reduces aliasing but also reduces image detail.
2. A small blur (small PSF) keeps details sharp but may increase aliasing.
3. Therefore, there is a balance between sharpness and image errors.
6. How PSF is Measured
1. By capturing a very small light source (like a pinhole).
2. By using special calibration patterns.
3. By analyzing the camera’s response to known inputs.
[Link] sensing pipeline

1. Scene Radiance
 Light comes from objects in the real world.
 It contains brightness, color, and details.
 This light enters the camera.
2. Optics (Lens System)
 The lens focuses the light onto the sensor.
 Controls image sharpness and viewing area.
 Better lens gives clearer images.
3. Aperture
 A small opening inside the lens.
 Controls how much light enters the camera.
 Large aperture → more light, blurry background.
 Small aperture → less light, more area in focus.
4. Shutter
 Controls how long light reaches the sensor.
 Fast shutter → freezes motion.
 Slow shutter → causes motion blur.
5. Sensor (CCD / CMOS)
 Captures light and converts it into electrical signals.
 Each pixel records light intensity.
6. Gain (ISO)
 Increases brightness of the signal.
 High ISO → brighter image but more noise.
 Low ISO → cleaner image but needs more light.
7. ADC (Analog to Digital Converter)
 Converts electrical signals into digital values.
 Creates digital image data.
8. RAW Output
 Unprocessed image data from the sensor.
 Contains maximum details.
9. Demosaicing
 Combines color data to form a full-color image.
 Needed because each pixel records only one color.
10. Denoising and Sharpening
 Removes noise from the image.
 Enhances edges and details.
11. White Balance
 Adjusts colors to look natural.
 Makes white objects appear white in any lighting.
12. Gamma / Curve Correction
 Adjusts brightness and contrast.
 Makes image suitable for display.
13. Compression
 Reduces file size.
 JPEG is commonly used.
14. JPEG Output
 Final processed image.
 Ready for viewing and sharing.

7. Photometric Image Formation (Lighting, Reflectance, and Shading)

1. Meaning of Photometric Image Formation
1. Photometric image formation explains how brightness and colors in an image are
produced.
2. Images are made of color and intensity values, not just shapes.
3. These values depend on light, object surfaces, and camera properties.
4. It describes how light interacts with objects and reaches the camera.
2. Lighting (Light Sources)
1. Images cannot be formed without light.
2. A scene must be illuminated by one or more light sources.
Types of light sources:
a) Point Light Source
1. Light comes from a single point (like a bulb or the Sun).
2. Has brightness (intensity) and color.
3. Light becomes weaker as distance from the object increases.
b) Area Light Source
1. Light comes from a larger surface (like a tube light or ceiling light).
2. Emits light in many directions.
3. Used to model soft and spread-out lighting.
Environment Lighting
1. Sometimes light comes from all directions (like outdoor light).
2. This can be represented using an environment map.
3. Reflectance and Shading (Light on Object Surfaces)
1. When light hits an object, it is reflected and scattered.
 2. The way light reflects depends on the material and surface shape.
3. This determines how bright or dark the object looks.
4. BRDF (Bi-directional ReflectanceDistribution Function)
1. BRDF describes how light reflects from a surface in different directions.
2. It depends on the incoming light direction and viewing direction.
3. Different materials reflect light differently.
5. Types of Reflection
a) Diffuse Reflection (Matte)
1. Light spreads equally in all directions.
2. Surface looks smooth and non-shiny.
b) Specular Reflection (Shiny)
1. Light reflects in a specific direction like a mirror.
2. Produces bright highlights on shiny objects.
c) Ambient Light
1. Background light from surroundings.
2. Lights up objects even without direct light.