Abstract:Hashing, as an efficient retrieval strategy, has been extensively studied in the image retrieval community. It aims to map the original high-dimensional features into compact binary codes, which dramatically reduces the cost of computing similarity with economic memory consumption. In general, most existing algorithms contain two stages: projecting high-dimensional descriptors into low-dimensional features, and encoding the low-dimensional features as binary strings. Although these two-stage approaches are intuitive and effective, certain beneficial information to binary encoding is inevitably discarded in the dimensionality reduction, which tends to lossy encoding. In this paper, we propose a novel hash algorithm where integrate subspace reconstruction and binary quantization into a unified framework and compact hash codes can be generated directly from high-dimensional features. More specifically, the proposed method discovers a latent subspace from the original feature space which is suitable for hash learning, thus alleviating the heavy information loss caused by traditional dimensionality reduction strategy. Meanwhile, thanks to the excellent property of rotational invariance in Euclidean space, the proposed method avoid the challenge of directly optimizing the binary matrix. Specially, our approach essentially is built on l2 estimator obeying the unsupervised paradigm, without involving any labeled data. Furthermore, we design an effective iterative alternating strategy to optimize the proposed model, so as to generate compact binary codes for the subsequent image matching. Extensive experiments on public benchmarks demonstrate that our method outperforms the compared method by a significant margin.

Hamming embedding with fragile bits for image search

Binary Code Reranking Method with Weighted Hamming Distance

Improved binary feature matching through fusion of hamming distance and fragile bit weight.

Binary Code Reranking Method Based on Bit Importance

Binary Feature from Intensity Quantization and Weakly Spatial Contextual Coding for Image Search.

Binary SIFT: towards efficient feature matching verification for image search

Coupled Binary Embedding for Large-Scale Image Retrieval

Triple-Bit Quantization with Asymmetric Distance for Image Content Security

Lost in Binarization

A Novel Binary Feature Descriptor for Accelerated Robust Matching

A Fuzzy Brute Force Matching Method for Binary Image Features

A Unified Binary Embedding Framework for Image Retrieval

Robust and Index-Compatible Deep Hashing for Accurate and Fast Image Retrieval.

Lost in binarization: query-adaptive ranking for similar image search with compact codes.

Learning Weighted Hamming Distance for Binary Descriptors

Improving feature matching strategies for efficient image retrieval

Circumferential Binary Feature Extraction and Matching Search Algorithms

Binary Code Ranking with Weighted Hamming Distance

Deep Forest Hashing for Image Retrieval

Matching binary feature search algorithm of bitmap locality sensitive hashing

On Aggregation of Unsupervised Deep Binary Descriptor with Weak Bits