Learning to Hash With Optimized Anchor Embedding for Scalable Retrieval.

Area of Science:

• Computer Vision
• Machine Learning
• Data Science

Background:

• Sparse representation and image hashing are key for data representation and retrieval.
• Sparse hashing (SH) methods combine these for scalable image retrieval, embedding high-dimensional features into low-dimensional Hamming spaces.
• Current SH methods primarily focus on optimizing sparse representations, often underestimating the importance of anchor sets.

Purpose of the Study:

• To propose a novel sparse hashing method that optimizes anchor set integration for improved feature embedding and binarization.
• To enhance the performance of sparse hashing by addressing the underestimation of anchor set importance in prior art.
• To develop an efficient optimization framework for the proposed method.

Main Methods:

• Proposed Sparse Hashing with Optimized Anchor Embedding (SHOA) method.
• Optimized anchor integration by pushing anchors away from the axis while preserving relative positions.
• Formulated the optimization as an orthogonality constrained maximization problem.
• Developed an efficient and novel optimization framework.

Main Results:

• Extensive experiments conducted on five benchmark image datasets.
• Demonstrated superior performance of the proposed SHOA method compared to several state-of-the-art related methods.
• SHOA effectively embeds and binarizes image features, generating similar hash codes for neighboring features.

Conclusions:

• The anchor set is a crucial component in sparse hashing methods, impacting feature embedding and retrieval performance.
• The proposed SHOA method offers a significant advancement in sparse hashing by optimizing anchor integration.
• SHOA provides a promising approach for scalable and efficient image retrieval.