Extension of Image and Video Fingerprinting Algorithm from Grayscale to Color domain

The Multimedia Information Management and Security (MIMS) group in UBC has been actively involved in developing multimedia fingerprinting algorithms that can assist content-providers with finding copies of their assets such as videos or images in online databases. Almost all of the current image and video fingerprinting algorithms are designed for grayscale images and videos. In other words, the original multimedia is first converted from color to grayscale and then its fingerprint is extracted for comparison purposes. While this pre-processing operation results in reducing the size of the input data, useful information is lost.

In this project, our goal is to extend the current image and video fingerprinting algorithms developed in the MIMS group from grayscale to color domain. For this purpose, we intend to use the theory of multiple classifier systems. Here, one classifier [i.e., a fingerprinting system] is first designed for each color domain (e.g., red, blue and green in an RGB space) and then the outputs of these fingerprinting systems are combined.

During this project, the students will be working closely with the researchers from the MIMS group to investigate various information fusion scenarios, investigate how the choice of the color domain (e.g., RGB vs. HSV) affects the performance and investigate the trade-off between the performance and the computational complexity.

Faculty Supervisor:

Dr. Rabab K. Ward

Student:

Gokul Raghuraman

Partner:

Discipline:

Engineering - computer / electrical

Sector:

Digital media

University:

University of British Columbia

Program: