Image Processing and Feature Extraction Review

A set of 20 vocabulary flashcards covering image processing, Sobel and Gaussian filtering, feature extraction, and feature scaling concepts.

Sobel Kernels

Two separate kernels ($\text{Sobel}_x$ and $\text{Sobel}_y$) containing negative and positive values that calculate the gradient to detect and highlight edges, requiring both image and kernel arrays to use floats.

Gaussian Kernel

An image processing filter designed to smooth an image, in contrast to Sobel kernels which detect and highlight edges.

x-Gradient Image

An image representing the gradient in the $x$ direction that observes variation from left to right while ignoring $y$-scaled variation.

y-Gradient Image

An image representing the gradient in the $y$ direction that observes variation from top to bottom while ignoring $x$-scaled variation.

Gradient

A measurement of the slope of pixel values in a given direction.

Threshold

The minimum value that a pixel needs to reach to be classified as an edge, such as converting any pixel value above $50$ to $255$ to form a white edge.

High Threshold Consequence

An outcome where the threshold value is set too high, causing some real edges to be left out because not enough pixels qualify as edges.

Low Threshold Consequence

An outcome where the threshold value is set too low, causing fainter outlines that might not be real edges to be caught because too many pixels qualify as edges.

Edge Detection

The process of outlining main structures and shapes in an image to facilitate image analysis and object classification.

Color Image Features

A set of $6$ HSV statistics (mean and standard deviation) extracted from a color image and returned in a list.

Grayscale Image Features

A set of $3$ geometric features extracted from a grayscale image: lines, circles, and triangles.

Circle Detection Filter

The Gaussian filter used to identify and detect circles in an image.

Triangle Detection Filters

The combination of both the Gaussian filter and the Sobel filter used to detect triangles.

Line Counting Filter

The Sobel filter used to detect and count lines in an image.

Metadata CSV File

A file that stores the file path and class label for each image in the dataset.

Feature Extraction

The process of transforming high-detail pixel data into numerical values that can be fed into machine learning models for image classification.

Features CSV File

An output file with $11$ columns containing the $9$ collected features, saving the processed dataset to prevent repetitive and computationally intensive calculations.

Feature Scaling

A standardization method that standardizes each feature by subtracting its mean and dividing by its standard deviation so that all features share a similar scale without altering their relationships.

Feature Space

A $2\text{D}$ or $3\text{D}$ coordinate space where each axis represents a feature and each image is plotted as a data point to reveal class clusters and separation.

Binary Shape Features

Features indicating the presence of shapes (such as triangles and circles) that only have values of $0$ or $1$, causing their feature plots to appear as parallel lines rather than continuous distributions.