基于E2LSH的近似重复图像检测

快速准确地检测出近似重复图像对于版权侵犯检测和检索结果冗余去除具有重要的意义。针对传统方法检测效率低的问题，提出了一种基于E2LSH（Exact Euclidean Locality Sensitive Hashing）的近似重复图像检测新方法。首先，利用E2LSH进行快速过滤，过滤掉大部分非近似重复图像对；然后，对于过滤后剩余的少量图像对，再采用视觉词袋法BoVW（Bag of Visual Words）进一步去除错误的匹配，从而实现近似重复图像检测。实验结果表明，新方法在不明显降低性能的前提下大大加快了近似重复图像检测的速度，在检测性能与速度之间取得了较好的平衡。

Near-Duplicate Image Detection Based on E2LSH

Detecting near-duplicate images efficiently and accurately is very important for copyright infringement detection and redundancy removal in retrieval results. In order to solve the low efficiency of the traditional methods, a new near-duplicate image detection method based on E2LSH is proposed. Firstly, E2LSH is used to filter most of the non near-duplicate image pairs. Then, a pool of vastly reduced image pairs is further invested by BoVW for removing the wrong matches, thus realizing the near-duplicate image detection. Experimental results show that the novel method significantly speed up the detection without apparent degradation in performance and achieve a good balance between performance and efficiency.