Content-Based Feature Extraction: Color Averaging
Rik Kamal Kumar Das · 2020
The color of an image is a vital feature to design meaningful descriptors for content-based image classification (CBIC). Color can be resourcefully utilized for designing feature vectors resulting in categorization of image data with high accuracy. This chapter considers block truncation coding (BTC) to differentiate a given image into three fundamental constituent colors. Further, it has demonstrated two different techniques—namely, feature extraction using BTC with color clumps and feature extraction using sorted block truncation coding ( S BTC). Guidelines for implementing the techniques in MATLAB® are also provided along with the description. Extracted features using these two techniques are compared with each other using diverse classification metrics and are evaluated for statistical significance in enhancing classification performances.