Scholarlink

<p>Defocus blur detection (DBD) plays a pivotal role in computer vision, serving as a fundamental step to enhance the performance of various downstream applications, such as image refocusing, depth estimation, and saliency detection. Despite recent advancements, existing methods often struggle in complex scenes with homogeneous regions, subtle blur transitions, and cluttered backgrounds. In this paper, we develop a novel approach (MSU-Mamba) to combine a multi-scale feature extraction with state-space modeling for boosting defocus blur detection. To do so, we develop a multi-scale fusion (MSF) block to integrate long-range dependency features across multiple scales to enhance feature representation. Moreover, in our MSF block, we devise a cross-scale token scanning mechanism into the original Mamba to better distinguish blurred and sharp regions. Comprehensive experiments conducted on benchmark datasets show that our MSU-Mamba outperforms state-of-the-art methods in terms of F-measure and MAE. The results validate our approach as a promising solution to the challenges of defocus blur detection and its application to downstream tasks.</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

Msu-mamba: multi-scale defocus blur detection using cross-scale fusion and state-space models

  • 期刊论文

  • 发表日期：

    2025年3月11日

查看全文
  • Xijun Wang,
  • Xin Zhou,
  • Yi Wang,
  • Songto Zeng,
  • Xinyu Liu,
  • Haobo Shen,
  • Song Fei,
  • Lei Zhu

摘要

Defocus blur detection (DBD) plays a pivotal role in computer vision, serving as a fundamental step to enhance the performance of various downstream applications, such as image refocusing, depth estimation, and saliency detection. Despite recent advancements, existing methods often struggle in complex scenes with homogeneous regions, subtle blur transitions, and cluttered backgrounds. In this paper, we develop a novel approach (MSU-Mamba) to combine a multi-scale feature extraction with state-space modeling for boosting defocus blur detection. To do so, we develop a multi-scale fusion (MSF) block to integrate long-range dependency features across multiple scales to enhance feature representation. Moreover, in our MSF block, we devise a cross-scale token scanning mechanism into the original Mamba to better distinguish blurred and sharp regions. Comprehensive experiments conducted on benchmark datasets show that our MSU-Mamba outperforms state-of-the-art methods in terms of F-measure and MAE. The results validate our approach as a promising solution to the challenges of defocus blur detection and its application to downstream tasks.