A Fast Algorithm for Near-Duplicate Image Detection

Abstract

Due to the popularity of digital image devices, the number of photos has increased dramatically. Because there are many photos with similar contents, the technology of detecting similar image is beneficial to the field of data storage, image retrieval and information security. This paper proposes an improved algorithm for near-duplicate image detection that not only utilizes the global feature of the color histogram, but also take into account of local complexity based on Entropy. It improves the accuracy of near-duplicate image detection without the increase of computing greatly. The experimental results on the image datasets indicate the effectiveness of the proposed algorithm.