天津大学学报
天津大學學報
천진대학학보
JOURNAL OF TIANJIN UNIVERSITY SCIENCE AND TECHNOLOGY
2014年
9期
771-777
,共7页
高贤君%万幼川%何培培%杨元维
高賢君%萬幼川%何培培%楊元維
고현군%만유천%하배배%양원유
阴影去除%综合补偿模型%云阴影%航空影像%高分辨率
陰影去除%綜閤補償模型%雲陰影%航空影像%高分辨率
음영거제%종합보상모형%운음영%항공영상%고분변솔
shadow removal%synthesized compensation model%cloud shadow%aerial image%high resolution
云阴影的存在严重影响了高分辨率航空遥感影像的视觉效果和影像质量。为去除云阴影,提出了一种无需人工参与的自动处理方法。首先采用光谱特征阈值法初步检测云阴影,再结合形态学方法进行后处理,提取完整的云阴影区域;阴影去除时,设计了一种利用非阴影区域信息、阴影整体区域信息、阴影像素局部窗口及像素本身信息的综合补偿模型,所需参数均可自动获取,进而实现阴影的自动去除。实验表明,区域级、窗口级和像素级多层次信息的引入,使模型的细节敏感度增强,提升了对比度,能灵活根据复杂的云阴影遮挡情况对阴影进行合理提升,有效解决了部分补偿方法存在的补偿效果不均匀的问题,阴影中地物的色彩与纹理等信息均得到更真实的再现。
雲陰影的存在嚴重影響瞭高分辨率航空遙感影像的視覺效果和影像質量。為去除雲陰影,提齣瞭一種無需人工參與的自動處理方法。首先採用光譜特徵閾值法初步檢測雲陰影,再結閤形態學方法進行後處理,提取完整的雲陰影區域;陰影去除時,設計瞭一種利用非陰影區域信息、陰影整體區域信息、陰影像素跼部窗口及像素本身信息的綜閤補償模型,所需參數均可自動穫取,進而實現陰影的自動去除。實驗錶明,區域級、窗口級和像素級多層次信息的引入,使模型的細節敏感度增彊,提升瞭對比度,能靈活根據複雜的雲陰影遮擋情況對陰影進行閤理提升,有效解決瞭部分補償方法存在的補償效果不均勻的問題,陰影中地物的色綵與紋理等信息均得到更真實的再現。
운음영적존재엄중영향료고분변솔항공요감영상적시각효과화영상질량。위거제운음영,제출료일충무수인공삼여적자동처리방법。수선채용광보특정역치법초보검측운음영,재결합형태학방법진행후처리,제취완정적운음영구역;음영거제시,설계료일충이용비음영구역신식、음영정체구역신식、음영상소국부창구급상소본신신식적종합보상모형,소수삼수균가자동획취,진이실현음영적자동거제。실험표명,구역급、창구급화상소급다층차신식적인입,사모형적세절민감도증강,제승료대비도,능령활근거복잡적운음영차당정황대음영진행합리제승,유효해결료부분보상방법존재적보상효과불균균적문제,음영중지물적색채여문리등신식균득도경진실적재현。
The existence of cloud shadow has a serious influence on the visual effect and image quality of aerial im-agery. In order to remove cloud shadow,an automatic method without any manual help was proposed. Firstly, according to the spectral characteristics of cloud shadow,the initial cloud shadow could be detected by spectral sig-nature with relative automatic threshold. Then the detection result would be post-processed by mathematical morphol-ogy to extract the complete cloud shadow area. Next,cloud shadow would be removed by a newly designed synthe-sized compensation model which utilizes the information of different areas including the whole shadow area and no-shadow area,a pixel-centered window area and the pixel itself. All the parameters in the model could be automati-cally acquired according to the information from aforementioned areas. Experimental results show that the use of dif-ferent information from whole area,window area and single pixel improves the ability of recognizing details and differences between pixels. In consequence,the contrast is promoted remarkably. Furthermore,the model is more flexible to deal with complex information shaded by cloud shadow,which resolves the problem existing in other methods that shadow areas are compensated unreasonably. The information of objects in shadow area including color and texture is restored more similarly to the truth.