Feature hashing

About: Feature hashing is a research topic. Over the lifetime, 993 publications have been published within this topic receiving 51462 citations.

Learning Independent, Diverse Binary Hash Functions: Pruning and Locality

Abstract: Information retrieval in large databases of complex objects, such as images, audio or documents, requires approximate search algorithms in practice, in order to return semantically similar objects to a given query in a reasonabletime. One practical approach is supervised binary hashing, where each object is mapped onto a small binary vector so that Hamming distances approximate semantic similarities, and the search is done in the binary space more efficiently. Much work has focused on designing objective functions and optimization algorithms for learning b-bit hash functions from a dataset. Recent work has shown that comparable or better results can be obtained by training b hash functions independently from each other and making them cooperate by introducing diversity with ensemble learning techniques. We show that this can be further improved by two techniques: pruning an ensemble of hash functions, and learning local hash functions. We show how it is possible to train our improved algorithms in datasets orders of magnitude larger than those used by most works on supervised binary hashing.

Local Feature Binary Coding for Approximate Nearest Neighbor Search

Abstract: The potential value of hashing techniques has led to it becoming one of the most active research areas in computer vision and multimedia. However, most existing hashing methods for image search and retrieval are based on global representations, e.g., GIST, which lack the analysis of the intrinsic geometric property of local features and heavily limit the effectiveness of the hash code. In this paper, we propose a novel supervised hashing method called Local Feature Binary Coding (LFBC) for projecting local feature descriptors from a high-dimensional feature space to a lower-dimensional Hamming space via compact bilinear projections rather than a single large projection matrix. LFBC takes the matrix expression of local features as input and preserves the feature-to-feature and image-to-class structures simultaneously. Experimental results on challenging datasets including Caltech-256, SUN397 and NUS-WIDE demonstrate the superiority of LFBC compared with state-of-the-art hashing methods.

Perceptual hash function for scalable video

Abstract: Perceptual hash functions are important for video authentication based on digital signature verifying the originality and integrity of videos. They derive hashes from the perceptual contents of the videos and are robust against the common content-preserving operations on the videos. The advancements in the field of scalable video coding call for efficient hash functions that are also robust against the temporal, spatial and bit rate scalability features of the these coding schemes. This paper presents a new algorithm to extract hashes of scalably coded videos using the 3D discrete wavelet transform. A hash of a video is computed at the group-of-frames level from the spatio-temporal low-pass bands of the wavelet-transformed groups-of-frames. For each group-of-frames, the spatio-temporal low-pass band is divided into perceptual blocks and a hash is derived from the cumulative averages of their averages. Experimental results demonstrate the robustness of the hash function against the scalability features and the common content-preserving operations as well as the sensitivity to the various types of content differences. Two critical properties of the hash function, diffusion and confusion, are also examined.

Interactive learning of multiple attribute hash table for fast 3D object recognition

Abstract: Multiple-attribute hashing is now considered to be a powerful approach for the recognition and localization of 3D objects on the basis of their invariant properties. In the systems developed to date, the hash tables must be created by the system developer/spl minus/an onerous task especially when the number of attributes is large, which is the case in systems that use both geometric and nongeometric attributes. The authors show how the tools of decision trees can be used for the automatic construction of hash tables. Their decision tree framework is based on a hybrid method that uses both qualitative attributes, such as the shape of a surface, and quantitative attributes such as color, dihedral angles, etc. In the system proposed the system developer shows objects to a vision system and, in an interactive mode, tells the system the model identities of the various segmented regions, etc. Subsequently, the decision tree based framework learns the structure of the hash table. >

Fast Image Search with Locality-Sensitive Hashing and Homogeneous Kernels Map

Abstract: Fast image search with efficient additive kernels and kernel locality-sensitive hashing has been proposed. As to hold the kernel functions, recent work has probed methods to create locality-sensitive hashing, which guarantee our approach's linear time; however existing methods still do not solve the problem of locality-sensitive hashing (LSH) algorithm and indirectly sacrifice the loss in accuracy of search results in order to allow fast queries. To improve the search accuracy, we show how to apply explicit feature maps into the homogeneous kernels, which help in feature transformation and combine it with kernel locality-sensitive hashing. We prove our method on several large datasets and illustrate that it improves the accuracy relative to commonly used methods and make the task of object classification and, content-based retrieval more fast and accurate.