Celebrity face image retrieval using multiple features

Abstract

Large scale face image retrieval is a hot topic in the field of internet retrieval. There exist a number of interesting applications of face image processing, such as hair-style design. In this paper, we propose a content-based face image retrieval system aiming at finding similar photos of celebrities to a user input image using a novel fusion of features and evaluation of results. After image preprocessing such as cropping facial parts and feature extraction with some exoteric methods, an algorithm we propose remolds the traditional features based on their statistic characteristics. The remolded features are then fused to a novel image representation to retrieve face images more effectively. A large number of experiments based on the dataset collected on the Internet demonstrate the good performance of our method in Mean Average Precision (MAP).