Topic
Locality-sensitive hashing
About: Locality-sensitive hashing is a research topic. Over the lifetime, 1894 publications have been published within this topic receiving 69362 citations.
Papers published on a yearly basis
Papers
More filters
TL;DR: An effective and efficient binary coding framework which enjoys the merits of lookup-based and hamming-based methods is proposed, and the relative error is proposed to preserve the relative sensitivity.
12 citations
22 Jun 2015
TL;DR: This work proposes to improve LSH by incorporating two elements - supervised hash bit selection and multi-scale feature representation and shows that the proposed supervision method is effective and the performance increases with the size of the hash bit pool.
Abstract: LSH is a popular framework to generate compact representations of multimedia data, which can be used for content based search. However, the performance of LSH is limited by its unsupervised nature and the underlying feature scale. In this work, we propose to improve LSH by incorporating two elements - supervised hash bit selection and multi-scale feature representation. First, a feature vector is represented by multiple scales. At each scale, the feature vector is divided into segments. The size of a segment is decreased gradually to make the representation correspond to a coarse-to-fine view of the feature. Then each segment is hashed to generate more bits than the target hash length. Finally the best ones are selected from the hash bit pool according to the notion of bit reliability, which is estimated by bit-level hypothesis testing. Extensive experiments have been performed to validate the proposal in two applications: near-duplicate image detection and approximate feature distance estimation. We first demonstrate that the feature scale can influence performance, which is often a neglected factor. Then we show that the proposed supervision method is effective. In particular, the performance increases with the size of the hash bit pool. Finally, the two elements are put together. The integrated scheme exhibits further improved performance.
12 citations
TL;DR: This paper proposes a novel approach to automatically detect near duplicate images based on visual word model using SIFT descriptors to represent image visual content and presents a local feature based image similarity estimating method by computing histogram distance.
Abstract: In recent years, near duplicate image detecting becomes one of the most important problems in image retrieval, and it is widely used in many application fields, such as copyright violations and detecting forged images. Therefore, in this paper, we propose a novel approach to automatically detect near duplicate images based on visual word model. SIFT descriptors are utilized to represent image visual content which is an effective method in computer vision research field to detect local features of images. Afterwards, we cluster the SIFT features of a given image into several clusters by the K-means algorithm. The centroid of each cluster is regarded as a visual word, and all the centroids are used to construct the visual word vocabulary. To reduce the time cost of near duplicate image detecting process, locality sensitive hashing is utilized to map high-dimensional visual features into low-dimensional hash bucket space, and then the image visual features are converted to a histogram. Next, for a pair of images, we present a local feature based image similarity estimating method by computing histogram distance, and then near duplicate images can be detected. Finally, a series of experiments are constructed to make performance evaluation, and related analyses about experimental results are also given
12 citations
01 Jan 2016
TL;DR: A new matching method is proposed which can reduce the detection time and enhance the accuracy of detection as well and the experimental results shows that the processing time has been reduced to 10% and the detection accuracy has been enhanced as well.
Abstract: Digital images are a main source of information in our modern digital era. However, the easiness of manipulating digital images using simple user-friendly software makes the credibility of images questionable. Copy-Move is one of the most common image forgery types, where a region of an image is copied and pasted into another location of the same image. Such a forgery is simple to achieve but hard to be detected as the pasted region shares the same characteristics with the image. Although plenty of algorithms have been proposed to tackle the copy-move detection problem, a fast and reliable copy-move detection algorithm is not achieved yet. In this paper, a new matching method is proposed which can reduce the detection time and enhance the accuracy of detection as well. Such enhancement is done by clustering image blocks into clusters, and searching for identical blocks within each cluster instead of all image blocks. For that purpose, k-means clustering is used to cluster the image blocks then Locality Sensitive Hashing (LSH) method is used to match the blocks based on Zernike moments. The experimental results shows that the processing time has been reduced to 10% and the detection accuracy has been enhanced as well.
12 citations
TL;DR: An approximate neighborhood queries method is presented for the computation of high dimensional data, in which, the locality-sensitive hashing (LSH) is used to reduce the computational complexity of the adaptive mean shift algorithm.
11 citations