Scholarlink

<p>Developing a video retrieval framework in multimedia management is a main challenge due to the massive growth of video content on the internet. A major drawback of video retrieval is its long search response time and low accuracy. To tackle these issues, this paper introduces a novel deep learning-based Multimedia video retrieval system (DL-MVR) to minimize the search response time with high accuracy. The collected video is initially converted into key frames and pre-processed with contrast adaptive histogram equalization to remove noise artifacts thereby improving image quality. After pre-processing, the images are fed to Efficient Net to extract patch features. Finally, to retrieve the similar video, matching is done using may fly optimization (MFO), that compares the query frame features to the video database. Several performance metrics are analysed to measure the effectiveness of the proposed strategy in terms of accuracy and response time. Experimental results indicate that the proposed system has a search response time of 0.71&#xa0;s, which is lower than existing methods. The proposed DL-MVR method achieves 99.26% of accuracy. The proposed method improves the overall accuracy by 9.32%, 22.04%, and 19.40% which is better than CNN-AlexNet (convolutional neural network), Pyramid regional graph network and CBVR respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel deep learning based multimedia video retrieval framework using may fly optimization

  • 期刊论文

  • 发表日期：

    2025年6月8日

查看全文
  • Ruchi Kaushik,
  • S. Pradeep,
  • Manjula Shanbhog,
  • A. Tamizhselvi,
  • Jayashree Jagdale,
  • Rashima Mahajan

摘要

Developing a video retrieval framework in multimedia management is a main challenge due to the massive growth of video content on the internet. A major drawback of video retrieval is its long search response time and low accuracy. To tackle these issues, this paper introduces a novel deep learning-based Multimedia video retrieval system (DL-MVR) to minimize the search response time with high accuracy. The collected video is initially converted into key frames and pre-processed with contrast adaptive histogram equalization to remove noise artifacts thereby improving image quality. After pre-processing, the images are fed to Efficient Net to extract patch features. Finally, to retrieve the similar video, matching is done using may fly optimization (MFO), that compares the query frame features to the video database. Several performance metrics are analysed to measure the effectiveness of the proposed strategy in terms of accuracy and response time. Experimental results indicate that the proposed system has a search response time of 0.71 s, which is lower than existing methods. The proposed DL-MVR method achieves 99.26% of accuracy. The proposed method improves the overall accuracy by 9.32%, 22.04%, and 19.40% which is better than CNN-AlexNet (convolutional neural network), Pyramid regional graph network and CBVR respectively.