A Hybrid Approach for Large-Scale Image Classification

Abstract

Digital photos and images increase considerably in recent years. How to manage and access large-scale images efficiently and effectively becomes an important issue in big data era. The main techniques of image retrieval are typically divides into two categories: the content-based retrieval approach and the keyword-based retrieval approach. The content-based retrieval approach searches images with a similar query image. In other hand, the keyword-based retrieval approach searches images with keywords. For retrieving images by semantic keywords and avoiding manually annotation by users, high-accurate image classification is required. In this paper, we propose a hybrid framework combining classifiers, similarity matching and association rules for large-scale multi-label image classification. The results of experiments show that the combination model provides an effective image classification.