Scholarlink

<p>In the era of exploding video content, effective video summarization methods are crucial for retaining essential information while maintaining relevance and usefulness. This paper proposes a novel single-view video summarization system, Frame Selection via Similarity Matching (FSSim), which systematically selects frames, assesses their quality using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM), and removes duplicates. The FSSim approach employs hash coding with Dice similarity to extract visually distinct frames, generating a compact and informative video summary. Experiments on multiple datasets, including TVSum, YouTube, Lobby, and real-time videos, demonstrate a significant reduction in video length while preserving informative content. The proposed framework achieves a high summarization accuracy, with user ratings ranging from 8.6 to 9.2 for informativeness, visual appeal, and usefulness. The FSSim method outperforms other techniques, achieving an F-score of 85.6%, precision of 88.1%, and recall of 91.2%, indicating its effectiveness and efficiency for video summarization. By open sourcing the code and datasets, this study aims to foster reproducibility and further advancements in the field.</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

Systematic frame selection and quality assessment for efficient video summarization

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • Payal Kadam,
  • Deepali Vora

摘要

In the era of exploding video content, effective video summarization methods are crucial for retaining essential information while maintaining relevance and usefulness. This paper proposes a novel single-view video summarization system, Frame Selection via Similarity Matching (FSSim), which systematically selects frames, assesses their quality using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM), and removes duplicates. The FSSim approach employs hash coding with Dice similarity to extract visually distinct frames, generating a compact and informative video summary. Experiments on multiple datasets, including TVSum, YouTube, Lobby, and real-time videos, demonstrate a significant reduction in video length while preserving informative content. The proposed framework achieves a high summarization accuracy, with user ratings ranging from 8.6 to 9.2 for informativeness, visual appeal, and usefulness. The FSSim method outperforms other techniques, achieving an F-score of 85.6%, precision of 88.1%, and recall of 91.2%, indicating its effectiveness and efficiency for video summarization. By open sourcing the code and datasets, this study aims to foster reproducibility and further advancements in the field.