Glossary

70 terms that recur across the curriculum. Skim before starting; refer back as needed.

Adaptive

Using locally-computed thresholds instead of a global value

Introduced in “Thresholding.”

Affine

Linear transformation preserving parallel lines

Introduced in “Image Transformations.”

Baseline

Physical distance between the two cameras

Introduced in “Stereo Vision & Depth.”

Bilateral Filter

Edge-preserving smoothing using spatial and intensity similarity

Introduced in “Noise Reduction.”

Binary Image

An image with only two values: black (0) and white (255)

Introduced in “Thresholding.”

Blob

A region differing from its surroundings in brightness or color

Introduced in “Blob Detection.”

Blur

Smoothing effect that reduces detail and noise

Introduced in “Convolution.”

Canny

Multi-stage edge detection algorithm considered the gold standard

Introduced in “Edge Detection.”

Color Range

Upper and lower bounds defining a target color in HSV

Introduced in “Color Tracking.”

Color Space

A system for representing colors mathematically

Introduced in “Pixels & Color Spaces.”

Connected Components

Groups of pixels that are touching (8-connectivity or 4-connectivity)

Introduced in “Blob Detection.”

Contour

A curve joining continuous boundary points of same intensity

Introduced in “Contour Detection.”

Convex Hull

Smallest convex polygon enclosing a shape

Introduced in “Contour Detection.”

Convolution

Mathematical operation of sliding a kernel across an image

Introduced in “Convolution.”

Corner

A point where two edges meet, detectable from all directions

Introduced in “Corner Detection.”

Depth Map

An image where pixel values represent distance

Introduced in “Stereo Vision & Depth.”

Disparity

Horizontal pixel difference between left and right images

Introduced in “Stereo Vision & Depth.”

DLT

Direct Linear Transform - algorithm for computing homography from correspondences

Introduced in “Homography.”

DoG

Difference of Gaussians - fast approximation to blob detection

Introduced in “Blob Detection.”

Edge

A rapid change in image intensity marking a boundary

Introduced in “Edge Detection.”

Epipolar Line

Line in one image where a point's correspondence must lie

Introduced in “Epipolar Geometry.”

Epipole

Image point where the line between camera centers intersects

Introduced in “Epipolar Geometry.”

Essential Matrix

Fundamental matrix for calibrated cameras, encoding R and t

Introduced in “Epipolar Geometry.”

Extrinsic Parameters

Rotation R and translation t describing camera pose

Introduced in “Camera Matrix & Projection.”

Face Detection

Locating faces (bounding boxes) in images or video

Introduced in “Face Detection.”

Feature Extraction

The process of identifying meaningful patterns in image data

Introduced in “What is Computer Vision?.”

Feature Point

A distinctive, trackable location in an image

Introduced in “Corner Detection.”

Filter

An operation that modifies image pixels based on their neighbors

Introduced in “Convolution.”

Focal Length

Distance from camera center to image plane, determines field of view

Introduced in “The Pinhole Camera Model.”

Fundamental Matrix

3×3 matrix encoding the epipolar constraint between two views

Introduced in “Epipolar Geometry.”

Gaussian Blur

Smoothing using a bell-curve weighted average of neighbors

Introduced in “Noise Reduction.”

Gradient

The rate of change of intensity in a direction

Introduced in “Edge Detection.”

Grayscale

Single-channel image representing intensity only

Introduced in “Pixels & Color Spaces.”

Haar Cascade

A sequence of classifiers using Haar-like rectangular features

Introduced in “Face Detection.”

Harris

Classic corner detection algorithm using eigenvalue analysis

Introduced in “Corner Detection.”

Homogeneous Coordinates

Extended coordinates allowing projection via matrix multiplication

Introduced in “Camera Matrix & Projection.”

Homography

A 3×3 matrix mapping points between two planes

Introduced in “Homography.”

HSV

Hue-Saturation-Value color space, separating color from brightness

Appears in “Pixels & Color Spaces” and “Color Tracking.”

Integral Image

A data structure allowing O(1) computation of rectangular sums

Introduced in “Face Detection.”

Interpolation

Estimating pixel values between known samples

Introduced in “Image Transformations.”

Intrinsic Matrix

3×3 matrix K encoding focal length and principal point

Introduced in “Camera Matrix & Projection.”

Kernel

A small matrix of weights used in convolution

Introduced in “Convolution.”

Laplacian

Second derivative operator that responds to intensity extrema

Introduced in “Blob Detection.”

Median Filter

Noise reduction by replacing pixels with the median of neighbors

Introduced in “Noise Reduction.”

Moments

Statistical measures describing shape distribution

Introduced in “Contour Detection.”

Morphology

Operations like erosion/dilation that modify shape based on structure

Introduced in “Color Tracking.”

Noise

Random variations in pixel values due to sensor/transmission imperfections

Introduced in “Noise Reduction.”

Object Tracking

Following an object's position across video frames

Introduced in “Color Tracking.”

Otsu

Automatic threshold selection minimizing within-class variance

Introduced in “Thresholding.”

Perspective

Transformation simulating 3D viewpoint change

Introduced in “Image Transformations.”

Pinhole

A theoretical camera model where light passes through a single point

Introduced in “The Pinhole Camera Model.”

Pipeline

A sequence of processing stages transforming raw images into decisions

Introduced in “What is Computer Vision?.”

Pixel

Picture element - the smallest unit of a digital image

Introduced in “Pixels & Color Spaces.”

Pixels

The smallest unit of a digital image, containing color/intensity values

Introduced in “What is Computer Vision?.”

Point Correspondence

A pair of matching points in two images

Introduced in “Homography.”

Principal Point

The image point where the optical axis intersects the sensor

Introduced in “The Pinhole Camera Model.”

Projection

The mapping of 3D world points to 2D image coordinates

Introduced in “The Pinhole Camera Model.”

Projection Matrix

The 3×4 matrix P = K[R|t] mapping 3D to 2D

Introduced in “Camera Matrix & Projection.”

RANSAC

Random Sample Consensus - robust estimation ignoring outliers

Introduced in “Homography.”

Real-time Processing

Processing video frames fast enough for live interaction (typically 30+ fps)

Introduced in “What is Computer Vision?.”

RGB

Red-Green-Blue color model used by displays

Introduced in “Pixels & Color Spaces.”

Shape Matching

Comparing shapes using invariant descriptors

Introduced in “Contour Detection.”

Sharpen

Enhancement that increases edge contrast

Introduced in “Convolution.”

Sobel

A gradient-approximation operator using 3×3 kernels

Introduced in “Edge Detection.”

Stereo

Using two cameras to perceive depth

Introduced in “Stereo Vision & Depth.”

Structure Tensor

A matrix summarizing local gradient structure

Introduced in “Corner Detection.”

Threshold

A cutoff value for classifying pixels as edge or non-edge

Appears in “Edge Detection” and “Thresholding.”

Transformation Matrix

A matrix encoding how to map input to output coordinates

Introduced in “Image Transformations.”

Triangulation

Computing 3D position from two 2D observations

Introduced in “Stereo Vision & Depth.”

Viola-Jones

The 2001 framework enabling real-time face detection

Introduced in “Face Detection.”