Scholarlink

<p>Poor quality retinal fundus images often hinder accurate diagnosis of ophthalmological diseases. The issue of poor image quality is addressed by developing an adaptive Riesz fractional derivative (RFD) mask in the presented work. This novel RFD-based adaptive approach (RFDAA) selectively enhanced the image texture. The optimal fractional order is selected from 0.1 to 0.9 based on performance metrics such as average gradient (AG) and information entropy (IE). The robustness is confirmed by considering the high-resolution fundus and eye diseases classification databases using IE, AG, and measure of enhancement (EME). All performance metrics increased significantly in comparison to existing techniques, with minimum improvements of 0.3536, 18.3814, and 10.1378 in IE, AG, and EME, respectively. A paired t-test is performed to further validate the effectiveness of the proposed approach, which shows that the performance of RFDAA is statistically significant compared to existing approaches. RFDAA can enhance image contrast while preserving image details. The textural features extracted from RFD-enhanced images can assist in the automatic classification of retinal fundus images into different categories.</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

Adaptive Fractional Derivative-Based Enhancement of Retinal Fundus Images

  • 期刊论文

  • 发表日期：

    2026年5月15日

查看全文
  • Kanwarpreet Kaur,
  • Neeru Jindal,
  • Kulbir Singh

摘要

Poor quality retinal fundus images often hinder accurate diagnosis of ophthalmological diseases. The issue of poor image quality is addressed by developing an adaptive Riesz fractional derivative (RFD) mask in the presented work. This novel RFD-based adaptive approach (RFDAA) selectively enhanced the image texture. The optimal fractional order is selected from 0.1 to 0.9 based on performance metrics such as average gradient (AG) and information entropy (IE). The robustness is confirmed by considering the high-resolution fundus and eye diseases classification databases using IE, AG, and measure of enhancement (EME). All performance metrics increased significantly in comparison to existing techniques, with minimum improvements of 0.3536, 18.3814, and 10.1378 in IE, AG, and EME, respectively. A paired t-test is performed to further validate the effectiveness of the proposed approach, which shows that the performance of RFDAA is statistically significant compared to existing approaches. RFDAA can enhance image contrast while preserving image details. The textural features extracted from RFD-enhanced images can assist in the automatic classification of retinal fundus images into different categories.