• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International symposium on photoelectronic detection and imaging;ISPDI 2011 >Analysis of Image Restoration and Evaluation for Diffraction-Degraded Remote Sensing Image
【24h】

Analysis of Image Restoration and Evaluation for Diffraction-Degraded Remote Sensing Image

机译：衍射退化遥感图像的图像复原分析与评估

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

A ringing effect often occurs in restored images, and its typical feature is that there are Gibbs-like oscillations in the neighboring areas of an image with sharp gray scale variations. The existence of the ringing effect causes it to be difficult for the restored images to be subsequently processed and some image quality evaluation methods to be invalid. In this paper, we studied several image restoration methods for diffraction-degraded remote sensing image: a Wiener filtering algorithm that is a simple and rapid image restoration algorithm, which is especially suitable for images without noise and with accurate PSF estimation and a small degree of blurring; an RL (Richardson-Lucy) algorithm that can gradually improve the image definition with an increase in iterations but the ringing effect becomes more and more significant; and a TV (total variation) algorithm that is a normalization algorithm based on noise and ringing suppression. We used multiple parameters to evaluate the restored images, including BDQ (block difference quality), GMG (gray mean grads), LS (Laplacian operator sum), and LE (large entropy), for which reference image are not required, as well as PSNR (peak signal noise ratio), SSIM (structural similarity), GRM(Gradient Ringing Metric) and HVSWGM (weighted gradient metric based on human visual system), for which reference image are required. The results show that the HVSWGM method is insensitive to the ringing effect occurring in image restoration and the evaluation result is completely consistent with a subjective evaluation result with a human visual system, and that many non-reference methods fail completely in assessing restored image with ringing effect, and only BDQ method is able to conform to subjective evaluation method to some extent.
机译：振铃效应通常发生在恢复的图像中，并且其典型特征是在图像的相邻区域中会出现类似于吉布斯的振荡，且灰度级会急剧变化。振铃效应的存在使得恢复的图像难以随后被处理，并且某些图像质量评估方法变得无效。在本文中，我们研究了几种用于衍射退化遥感图像的图像恢复方法：一种维纳滤波算法，这是一种简单而快速的图像恢复算法，特别适用于无噪声，PSF估计准确且图像失真度小的图像。模糊RL（Richardson-Lucy）算法可以随着迭代次数的增加而逐渐改善图像清晰度，但振铃效果变得越来越重要; TV（总变化）算法是基于噪声和振铃抑制的归一化算法。我们使用了多个参数来评估恢复的图像，包括不需要参考图像的BDQ（块差质量），GMG（灰色平均灰度），LS（拉普拉斯算子和）和LE（大熵），以及需要参考图像的PSNR（峰值信号噪声比），SSIM（结构相似性），GRM（梯度振铃指标）和HVSWGM（基于人类视觉系统的加权梯度指标）。结果表明，HVSWGM方法对图像还原中出现的振铃效果不敏感，并且评估结果与人眼视觉系统的主观评估结果完全一致，并且许多非参考方法完全无法评估带有振铃的还原图像效果，并且只有BDQ方法能够在一定程度上符合主观评估方法。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号