Abstract:Feature expression and similarity metric are two crucial steps for content-based image retrieval. In this paper, an novel image retrieval approach based on semantic affinity is proposed through exploiting discriminative features and capturing the latent data manifold. Different from the traditional practice, firstly Axiomatic Fuzzy sets (AFS) is introduced to masterly embed the semantic into original image features for forming a new semantic feature space. Then different semantic descriptions are exploited for different images according real image attribute distribution, which effectively reflect the semantic difference between images. Finally, a semantic affinity metric is defined for image matching based on the similarities of semantic descriptions. It is noteworthy that this proposed metric consider not only the similarities between two images but also their their respective neighbors, which is totally different from the traditional distance metrics model only base on pair-wise distance of images. With accurate description and reflection of semantic affinity, the proposed method can further enhance the stability and robustness of similarity relationship of images and thereby leads to improvement of the retrieval performance significantly. Extensive experiments on benchmark databases verify that it can achieve a good performance in completing the face retrieval task based on pixel-level grayscale features and the natural image retrieval task based on CNNs depth features.

Two-Stage Texture Image Retrieval by a Combination of Semantic Feature and Visual Feature

Semantic Image Retrieval Based on Multiple-Instance Learning

SIFT-Based Image Retrieval Combining the Distance Measure of Global Image and Sub-Image

Image Retrieval Based on Fuzzy Semantic Relevance Matrix

Two-stage Content Based Image Retrieval Using Sparse Representation and Feature Fusion

Image Retrieval Based on Text and Semantic Relevance Analysis

ImageSaker : A Semantic-based Image Retrieval System Refining with Concept Model

A Method for Content-Based Image Retrieval with a Two-Stage Feature Matching

Real-Time Image Semantic Retrieval Based on VQ

AN APPROACH TO SEMANTICS-BASED IMAGE RETRIEVAL AND BROWSING

Research of Image Retrieval Method Based on Improved Feature

Content-Based Image Retrieval Using Iterative Search

A new image retrieval system supporting query by semantics and example

Semantic-based image retrieval algorithm using fuzzy support vector machine

Semantic Template: A Robust Approach To Support Content-Based Image Retrieval

An image semantic retrieval system design and realization

A Semantic Indexing Structure for Image Retrieval

Two Image Retrieval Methods Based on Support Vector Machines Semantic Classification

A novel image retrieval approach based on semantic affinity

Thesaurus-aided Approach for Image Browsing and Retrieval

A Retrieval Mechanism for Complex Similarity Queries in Image Databases