Get started
Work

Image retrieval method based on data mining and deep residual network

Year: 2025
Type: article
Abstract: To address the significant shortcomings of traditional image retrieval models in feature extraction and recognition accuracy, this study proposes a new method based on the improved residual network and attention mechanism. By constructing convolutional neural networks , the research aims to achieve multi-level automatic feature learning, thereby significantly improving the extraction efficiency of image information. To overcome the vanishing gradient problem of traditional residual networks in deep structures, an improved residual network structure was introduced in the study, and optimization was performed simultaneously by combining the spatial transformation network and the feature enhancement module. The experimental results showed that the improved model achieved an average retrieval accuracy of 81.1 % on three different image sets, which was 27 percentage points higher than that of the traditional model. Among the 5358 test images, the highest retrieval accuracy even reached 82.7 %. Furthermore, with the effective integration of the attention mechanism, the performance of the model under different peak signal-to-noise ratio and structural similarity conditions was also significantly improved, and the retrieval accuracy reached 92.4 % at 30 dB. These results indicate that the proposed improved model not only significantly improves the accuracy and real-time performance of image retrieval, but also optimizes the extraction ability of image features. This achievement has proposed an innovative solution for the field of image retrieval and demonstrated the broad application potential of data mining technology in image understanding.
DOI: 10.1016/j.sasc.2025.200331
Authors: Hongzhi Yuan, Wei Hu
Language: English
Open Access status: gold
Estimated APC paid: $1,870
OpenAlex ID: W4411276674
FWCI: 1.124
Cites: 33
Cited by: 3
Related to: 10