Health Indicators from Flickr Photos – A Color Based Image Analysis

Published in: Megaprojects: Building Infrastructure by Fostering Engineering Collaboration, Efficient and Effective Integration and Innovative Planning: Proceedings of the 10th Latin American and Caribbean Conference for Engineering and Technology
Date of Conference: July 23-25, 2012
Location of Conference: Panama City, Panama
Authors: Francisco Justo
Hari Kalva
Daniel Raviv
Refereed Paper: #245

Abstract

This paper presents the design and testing of a tool which uses color recognition from images taken from Flickr to predict if the food in that image is healthy or not. The tool basically transforms an image to a 64-color pallet and builds a color histogram for the image. A further reduction to 11 commonly found colors in photos of food is done to complete color analysis. The proportion corresponding to each color with respect to the size of the image was calculated. In addition, the tool has two filters which are used to identify if the image contain healthy food or not. The experiments showed that eliminating the unwanted objects from the images helps to improve the performance of the tool. The proposed tool is a first step in developing a platform for extracting health indicators from social signals such as Flickr photos.