Skip to content

nepiskopos/image-classification-textural-features

Repository files navigation

Image classification using textural features

This repository contains some MATLAB scripts with a simple implementation of image classification using properties of gray-level co-occurence matrix.

More specifically, I use Contrast, Homogeneity (Correlation) and Energy (Entropy) to classify grayscale textures that originate from 6 distinct classes, with the help of 3-Nearest Neighbors (3-NN) and Minimum Euclidean Distance classifiers.

The 3 original TIFF images/classes are located in the original-images directory.

Using these 6 images, I generate 24 images as training samples (4 from each class) by extracting parts of the original images.

I also generate 20 random images as test objects that are randomly selected from the classes and consist of parts of the original images.

The classification features consist of the first Contrast, Homogeneity and Energy.

Using these 3 features, I classify the random objects to the 6 classes, using the classifiers mentioned above.

To execute the code using MATLAB (GNU Octave is not supported due to the lack of co-occurence matrix related functions), naviagte to the directory with the scripts and run main.m

About

Pattern recognition image classification using textural features.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages