Scholarlink

In this paper, we investigate if the well-known Structural Similarity image quality measure (SSIM) can be improved by incorporating gradient information. We propose a simple gradient similarity measure which yields results similar to the canonical correlation method. Using the LIVE image database, we show that our proposed gradient-based SSIM exhibits improved performance for degraded images in the mid-to-low quality range.

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

Exploring the Use of Gradients in the Structural Similarity Image Quality Measure

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amelia Kunze,
  • Edward R. Vrscay

摘要

In this paper, we investigate if the well-known Structural Similarity image quality measure (SSIM) can be improved by incorporating gradient information. We propose a simple gradient similarity measure which yields results similar to the canonical correlation method. Using the LIVE image database, we show that our proposed gradient-based SSIM exhibits improved performance for degraded images in the mid-to-low quality range.