Threadz: Image-Matching Using Colors and Contrasts

Document Type

Project

Lead Author Type

CIS Masters Student

Advisors

Dr. Greg Wolffe, wolffe@gvsu.edu

Embargo Period

5-21-2013

Abstract

One of the objectives of the furniture industry is to help customers match fabric options with their home or office décor. In particular, the ability to automatically search through a fabric database by color would be valuable for designers and customers alike. However, text- based color descriptions are subjective: “red” and “blue” have too wide a range, and it is difficult to reach agreement on terms such as “pale” or “electric”. Therefore, the best option is to have the user present a direct visual representation of the color they desire.

The goal of this project was to develop a two-tiered application to solve the color communication problem. The first part is an Android application, which allows the user to deliver an image off a digital device via the gallery/device storage or the camera. The idea is that a user could bring in a digital image of their carpeting, walls or other furniture. An intelligent front end processes the image to extract meaningful color information. The second part of the solution implements a Tomcat-based web service that receives color information sent from the device and interfaces with a fabric database. Both tiers use OpenCV, an open-source real-time computer vision library, to implement search and comparison routines. The result is a solution that returns a selection of matching fabrics, including the use of similarity metrics to determine complementary colors.

This document is currently not available here.

Share

COinS