期刊文献+

基于低秩分解与像素置乱的图像去雾方法 被引量:3

Image Defogging Method Based on Low-Rank Decomposition and Pixel Scrambling
下载PDF
导出
摘要 针对浓雾场景下图像目标信息被严重遮挡,现有雾天图像清晰化算法难以取得较好去雾效果的问题,基于低秩分解并结合像素置乱提出一种新的图像去雾方法。根据低秩分解理论和散射介质成像模型,将雾天降质图像看作两部分的叠加:一部分是具有低秩特性的雾化背景,另一部分是具有高秩特性的清晰目标场景。由于目标场景本身具有局部相关性和非局部相似性而含有一定程度的低秩成分,直接进行低秩分解会导致一部分目标场景被当作雾化背景去除,因此对原始雾天图像进行像素置乱以破坏场景本身的相关性,同时雾化背景因其全局缓变特性仍保持低秩属性,从而在进行低秩分解时最大限度地保留场景信息。最后,将高秩成分进行像素归位,获得去雾后的复原场景。实验结果表明,与暗通道先验、DehazeNet等主流图像去雾方法相比,该方法针对O-HAZE数据集中浓雾图像的去雾具有更好的表现,在有效去除浓雾的同时,不会产生大面积色偏现象。 An image defogging method based on low-rank decomposition combined with pixel scrambling is proposed to address the problem of image target information being severely occluded in dense fog scenes and existing foggy image sharpening algorithms failing to achieve a good dehazing effect.The foggy degraded image is regarded as the sum of two parts,according to the low-rank decomposition theory and the scattering medium imaging model:one part is the foggy background with a low-rank attribute,and the other part is the clear target scene with a high-rank attribute.Because the target scene contains a certain number of low-rank components due to local correlation and non-local similarity,direct low-rank decomposition will result in a part of the target scene being removed as the foggy background.Therefore,pixel scrambling is performed on the original foggy image to destroy the correlation of the scene,and the foggy background still retains the low-rank attribute due to its global gradual change,allowing the scene information to be preserved to the greatest extent possible during low-rank decomposition.Finally,the restored scene after dehazing can be obtained by resetting the pixels in high-rank components.The experimental results show that this method performs better on dense foggy image dehazing and does not produce a large-area color cast while effectively removing dense fog.When compared to mainstream image dehazing methods such as dark channel prior and DehazeNet,this method reduces dehazing time while increasing the optimal Peak Signal-to-Noise Ratio(PSNR)for dense fog images in the O-HAZE dataset.
作者 王国栋 邵鹏 王国宇 刘少禹 张建涛 WANG Guodong;SHAO Peng;WANG Guoyu;LIU Shaoyu;ZHANG Jiantao(School of Electronic Engineering,Faculty of Information Science and Engineering,Ocean University of China,Qingdao,Shandong 266100,China)
出处 《计算机工程》 CAS CSCD 北大核心 2022年第12期212-217,共6页 Computer Engineering
基金 国家自然科学基金“浑浊水体中多照明模式下成像优化处理方法研究”(61571407)。
关键词 图像去雾 低秩分解 散射介质成像模型 像素置乱 像素归位 image defogging low-rank decomposition scattering medium imaging model pixel scrambling pixel resetting
  • 相关文献

参考文献7

二级参考文献89

  • 1陈明惠,王帆,张晨曦,李福刚,郑刚.基于压缩感知的频域OCT图像稀疏重构[J].光学精密工程,2020,28(1):189-199. 被引量:18
  • 2Gonzales Rafael C.数字图像处理[M].北京:电子工业出版社.2005.
  • 3章毓晋.图像工程(上册)[M].北京:清华大学出版社.2006.56-58.
  • 4JIANG Xu-dong. Linear subspace learning-based dimensionality reduction [ J ]. I EEE Signal Processing Magazine,2011,28 ( 2 ) : 16- 26.
  • 5DONOHO D L. Compressed sensing[ J]. IEEE Trans on Information Theory,2006,52 (4) : 1289-1306.
  • 6CANDES E J, MICHAEL W. An introduction to compressive sampling [J]. IEEE Signal Processing Magazine,2008,25(2) :21-30.
  • 7WRIGHT J, ALLEN Y, GANESH A, et al. Robust face recognition via sparse representation[ J]. IEEE Trans on Pattern Analysis and Machine Intelligence ,2009,31 ( 2 ) :210-227.
  • 8WRIGHT J, MA Yi, MAIRAL J, et al. Sparse representation for computer vision and pattern recognition [ J ]. Proceedings of the IEEE,2010, 98(6) :1031-1044.
  • 9WRIGHT J, GANESH A, RAO S, et al. Robust principal component analysis : exact recovery of corrupted low-rank matrices via convex optimization [ C ]//Proc of the 24th Annual Conference on Neural Information Processing Systems. 2009 : 2080-2088.
  • 10CANDES E J, LI Xiao-dong, MA Yi, et al. Robust principal component analysis? [J]. dournal of the AOM,2011,58(3) :1-37.

共引文献111

同被引文献29

引证文献3

二级引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部