👁️Computer Vision and Image Processing Unit 1 – Image Formation and Representation

Key Concepts

• Image formation involves capturing light from a scene and converting it into a digital representation
• Digital images are represented as a 2D matrix of pixels, each with a specific color or intensity value
• Color models (RGB, CMYK, HSV) define how colors are represented and combined in digital images
• Image sampling determines the spatial resolution of an image by measuring the number of pixels per unit area
• Quantization assigns discrete intensity values to each pixel, affecting the image's color depth and file size
• Image file formats (JPEG, PNG, TIFF) specify how image data is stored, compressed, and exchanged
• Image enhancement techniques (contrast adjustment, noise reduction, sharpening) improve visual quality and highlight features
• Practical applications of image processing include medical imaging, surveillance, remote sensing, and computer vision tasks

Image Formation Process

• Image formation captures light from a 3D scene and projects it onto a 2D plane
• The process involves a camera or imaging device with a lens and an image sensor (CCD or CMOS)
• Light rays from the scene pass through the lens and are focused onto the image sensor
• The lens controls the amount of light entering the camera through its aperture and focal length settings
  • Aperture size affects depth of field and exposure
  • Focal length determines the angle of view and magnification
• The image sensor consists of a grid of light-sensitive elements called photosites or pixels
• Each photosite converts the incoming light into an electrical signal proportional to the light intensity
• The electrical signals are then processed by the camera's image processing unit to generate a digital image

Digital Image Representation

• A digital image is a 2D matrix of picture elements called pixels
• Each pixel represents a small area of the image and has a specific color or intensity value
• The spatial resolution of an image is determined by the number of pixels in the matrix (width × height)
  • Higher resolution images have more pixels and can capture finer details
  • Lower resolution images have fewer pixels and may appear pixelated or blurry
• The color depth or bit depth of an image determines the number of possible color or intensity values for each pixel
  • Common color depths include 8-bit (256 values), 16-bit (65,536 values), and 24-bit (16.7 million values)
• Grayscale images have pixels with intensity values ranging from black (lowest) to white (highest)
• Color images have pixels with a combination of color components (red, green, blue) or color channels

Color Models and Spaces

• Color models define how colors are represented and combined in digital images
• The RGB (Red, Green, Blue) model is commonly used for display devices and digital cameras
  • Each pixel has three color components (red, green, blue) with values ranging from 0 to 255
  • Colors are created by combining different intensities of red, green, and blue light
• The CMYK (Cyan, Magenta, Yellow, Key/Black) model is used for printing and subtractive color mixing
• The HSV (Hue, Saturation, Value) model represents colors based on their hue, saturation, and brightness
  • Hue describes the dominant wavelength of the color (0-360 degrees on a color wheel)
  • Saturation indicates the purity or intensity of the color (0-100%)
  • Value represents the brightness or lightness of the color (0-100%)
• Other color spaces include Lab, YCbCr, and XYZ, which are used for specific applications or color management systems

Image Sampling and Quantization

• Image sampling is the process of measuring the spatial resolution of an image
• It determines the number of pixels per unit area (pixels per inch or dots per inch)
• Higher sampling rates result in higher spatial resolution and more detailed images
• Lower sampling rates lead to lower spatial resolution and potential loss of information
• Quantization is the process of assigning discrete intensity or color values to each pixel
• It maps the continuous range of light intensities to a finite set of discrete values
• The number of quantization levels determines the color depth or bit depth of the image
  • More quantization levels allow for a wider range of colors or intensities
  • Fewer quantization levels result in a smaller color palette and potential banding artifacts
• The trade-off between sampling rate and quantization levels affects image quality, file size, and processing requirements

Image File Formats

• Image file formats specify how image data is stored, compressed, and exchanged
• JPEG (Joint Photographic Experts Group) is a lossy compression format widely used for photographs
  • It achieves high compression ratios by discarding high-frequency information and using DCT-based compression
  • JPEG is suitable for images with smooth color transitions but may introduce compression artifacts at high compression levels
• PNG (Portable Network Graphics) is a lossless compression format that supports transparency
  • It uses a combination of filtering and DEFLATE compression to reduce file size without losing quality
  • PNG is suitable for images with sharp edges, text, or graphics but may result in larger file sizes compared to JPEG
• TIFF (Tagged Image File Format) is a flexible format that supports both lossy and lossless compression
  • It can store multiple images, layers, and metadata within a single file
  • TIFF is commonly used in professional photography and publishing workflows
• Other formats include GIF (Graphics Interchange Format), BMP (Bitmap Image File), and RAW (camera-specific formats)

Image Enhancement Techniques

• Image enhancement techniques improve the visual quality and interpretability of digital images
• Contrast enhancement adjusts the distribution of pixel intensities to increase the dynamic range and visibility of details
  • Histogram equalization redistributes pixel intensities to achieve a more uniform distribution
  • Contrast stretching expands the range of intensities to occupy the full available range
• Noise reduction techniques remove or suppress unwanted distortions and artifacts in the image
  • Gaussian filtering applies a weighted average to neighboring pixels to smooth out noise
  • Median filtering replaces each pixel with the median value of its local neighborhood
• Sharpening techniques emphasize edges and fine details in the image
  • Unsharp masking subtracts a blurred version of the image from the original to highlight edges
  • High-pass filtering accentuates high-frequency components and suppresses low-frequency components
• Color correction methods adjust the color balance, saturation, and hue of an image to achieve a desired appearance
• Image enhancement techniques can be applied globally to the entire image or locally to specific regions of interest

Practical Applications

• Medical imaging uses image processing techniques to visualize and analyze anatomical structures and pathologies
  • X-ray, CT, MRI, and ultrasound images are enhanced to improve diagnostic accuracy and treatment planning
• Surveillance and security systems rely on image processing for object detection, tracking, and recognition
  • Facial recognition, license plate detection, and anomaly detection are common applications
• Remote sensing and satellite imagery employ image processing for Earth observation and environmental monitoring
  • Multispectral and hyperspectral imaging techniques are used for land cover classification, vegetation analysis, and change detection
• Computer vision tasks such as object recognition, segmentation, and depth estimation heavily rely on image processing algorithms
  • Convolutional neural networks (CNNs) and deep learning models are trained on large datasets of images to learn hierarchical features and patterns
• Image processing is also applied in fields like astronomy (image stacking, artifact removal), forensics (image enhancement, forgery detection), and industrial inspection (defect detection, quality control)