Scholarlink

This paper introduces “DSFM Resnet,” (ResNet with Depth Separable convolution augmenting Fractal Dimension and MAE Loss) a novel denoising architecture specifically designed to enhance the quality of stone inscription images. The proposed method integrates depthwise separable convolution (DS), fractal dimension loss (FD), and mean absolute error (MAE) within the Resnet framework to achieve exceptional denoising performance. Extensive experiments conducted on a curated dataset of stone inscription images show that DSFM Resnet consistently outperforms other DN_Resnet variants, achieving an impressive PSNR of 39.29 and an SSIM of 0.9959. The model maintains its efficiency through the use of depthwise separable convolution, while the incorporation of fractal dimension loss allows for the capture of intricate patterns unique to stone inscriptions. The DSFM Resnet architecture has potential applications in stone inscription restoration, artifact preservation, and historical analysis, making it a compelling solution for enhancing quality and conserving historical artifacts.

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

DSFM Resnet: A Novel Denoising Architecture for Stone Inscription for Image Restoration

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • R. Archana,
  • P. S. Eliahim Jeevaraj

摘要

This paper introduces “DSFM Resnet,” (ResNet with Depth Separable convolution augmenting Fractal Dimension and MAE Loss) a novel denoising architecture specifically designed to enhance the quality of stone inscription images. The proposed method integrates depthwise separable convolution (DS), fractal dimension loss (FD), and mean absolute error (MAE) within the Resnet framework to achieve exceptional denoising performance. Extensive experiments conducted on a curated dataset of stone inscription images show that DSFM Resnet consistently outperforms other DN_Resnet variants, achieving an impressive PSNR of 39.29 and an SSIM of 0.9959. The model maintains its efficiency through the use of depthwise separable convolution, while the incorporation of fractal dimension loss allows for the capture of intricate patterns unique to stone inscriptions. The DSFM Resnet architecture has potential applications in stone inscription restoration, artifact preservation, and historical analysis, making it a compelling solution for enhancing quality and conserving historical artifacts.