中国科技论文
中國科技論文
중국과기논문
China Sciencepaper
2015年
20期
2405-2411
,共7页
原惠峰%杨风暴%张雷%牛涛
原惠峰%楊風暴%張雷%牛濤
원혜봉%양풍폭%장뢰%우도
图像融合%偏振图像%红外光强图像%性能比较
圖像融閤%偏振圖像%紅外光彊圖像%性能比較
도상융합%편진도상%홍외광강도상%성능비교
image fusion%polarization image%infrared light intensity image%performance comparison
现有红外偏振图像融合算法多采用单一固定的融合技术对源图像进行描述,在实际融合中无法全面描述、提取变化的源图像差异特征,不利于融合。针对这一问题,对比分析了多种图像融合技术,归纳总结了不同融合技术在融合红外偏振与光强图像时的特点,引入了梯形模糊数构建客观评价指标阈值,对多组图像的融合效果进行对比分析。研究表明:小波包变换(wavelet packet transform,WPT),双树复小波变换(dual-tree complex wavelet transform,DTCWT),非下采样剪切波变换(non-subsampled shearlet transform,NSST),非下采样轮廓波变换(non-subsampled contourlet transform,NSCT),主成分分析(princi-pal component analysis,PCA)等融合技术能较充分地融合源图像的信息;WPT、DTCWT、NSST、NSCT、PCA、顶帽变换(Top-hat)等融合技术较好地融合图像的边缘、轮廓;离散小波变换(discrete wavelet transform,DWT),多分辨率奇异值分解(multi-resolution singular value decomposition,MSVD),Top-hat 等融合技术的融合结果有明显的块效应、锯齿状边缘;当某一源图像提供的信息量较大时,PCA 相比其他融合技术更具有优势,但会丢失重要的局部特征。
現有紅外偏振圖像融閤算法多採用單一固定的融閤技術對源圖像進行描述,在實際融閤中無法全麵描述、提取變化的源圖像差異特徵,不利于融閤。針對這一問題,對比分析瞭多種圖像融閤技術,歸納總結瞭不同融閤技術在融閤紅外偏振與光彊圖像時的特點,引入瞭梯形模糊數構建客觀評價指標閾值,對多組圖像的融閤效果進行對比分析。研究錶明:小波包變換(wavelet packet transform,WPT),雙樹複小波變換(dual-tree complex wavelet transform,DTCWT),非下採樣剪切波變換(non-subsampled shearlet transform,NSST),非下採樣輪廓波變換(non-subsampled contourlet transform,NSCT),主成分分析(princi-pal component analysis,PCA)等融閤技術能較充分地融閤源圖像的信息;WPT、DTCWT、NSST、NSCT、PCA、頂帽變換(Top-hat)等融閤技術較好地融閤圖像的邊緣、輪廓;離散小波變換(discrete wavelet transform,DWT),多分辨率奇異值分解(multi-resolution singular value decomposition,MSVD),Top-hat 等融閤技術的融閤結果有明顯的塊效應、鋸齒狀邊緣;噹某一源圖像提供的信息量較大時,PCA 相比其他融閤技術更具有優勢,但會丟失重要的跼部特徵。
현유홍외편진도상융합산법다채용단일고정적융합기술대원도상진행묘술,재실제융합중무법전면묘술、제취변화적원도상차이특정,불리우융합。침대저일문제,대비분석료다충도상융합기술,귀납총결료불동융합기술재융합홍외편진여광강도상시적특점,인입료제형모호수구건객관평개지표역치,대다조도상적융합효과진행대비분석。연구표명:소파포변환(wavelet packet transform,WPT),쌍수복소파변환(dual-tree complex wavelet transform,DTCWT),비하채양전절파변환(non-subsampled shearlet transform,NSST),비하채양륜곽파변환(non-subsampled contourlet transform,NSCT),주성분분석(princi-pal component analysis,PCA)등융합기술능교충분지융합원도상적신식;WPT、DTCWT、NSST、NSCT、PCA、정모변환(Top-hat)등융합기술교호지융합도상적변연、륜곽;리산소파변환(discrete wavelet transform,DWT),다분변솔기이치분해(multi-resolution singular value decomposition,MSVD),Top-hat 등융합기술적융합결과유명현적괴효응、거치상변연;당모일원도상제공적신식량교대시,PCA 상비기타융합기술경구유우세,단회주실중요적국부특정。
Random features of source images cannot be fully described and extracted by using existing infrared polarization image fusion algorithm,in which the single fixed fusion technology is applied to describe the source images.Aiming at the problem,this paper comparatively analyzes fusion technology and the features of difference fusion technology in the fusion infrared polarization and light intensity image.The fuzzy number is introduced to construct an objective evaluation threshold and analyses multiple sets of image fusion effect.It is shown that WPT,DTCWT,NSST,NSCT and PCA can be fully integrated information of source im-age;WPT,DTCWT,NSST,NSCT,PCA and Top-hat can be better integrated in edge and contour of the image;the fusion re-sults of DWT,MSVD and Top-hat are obvious block effect and serrated edge.When a source image provides large amount of in-formation,PCA presents better than other fusion technology,but would lose important local features.