Introduction

The colorfulness of a rendered black and white image is an indication of the distortion introduced from image sampling and processing. Image metrics such as CIELAB and spatial-CIELAB (sCIELAB) provide a basis for identifying the relative colorfulness of an image, but there is no metric to quantify the distortion. Furthermore, it is difficult to simulate whether an image will appear colorful or not. Past work includes computing colorfulness based on statistical parameters of the pixel cloud along red-green and yellow-blue axes (Palus); a spatial extension to the CIELAB color metric (Zhang and Wandell); and a colorfulness metric meant to be used real time on video streams (Hasler and Süsstrunk). Our metric is designed to evaluate the colorfulness of an image based on behavioral data integrated with a variety of classifiers to gain insight into factors that impact a viewer’s perception of color distortion.

· View the written report as Adobe PDF format.

· View the PowerPoint presentation.

· Link to MATLAB code and data files.