SSIM分析和总结分析和总结.docxVIP

SSIM 指数是 Zhou Wang 在 2004 年提出的一个图像质量评价指标（Digital Object Identifier 10.1109/TIP.2003.819861）。因为该指标能较好的反映图像的视觉效果而受到广泛关注和使用。但现有代码都是 matlab 版本，为了方便 ENVI/IDL 用户使用，将该版本翻译为 IDL 格式。如下： 代码要求： 图像路径和文件名要手动输入 图像为 ENVI 标准格式即可 proaverage_SSIM ;define file names++++++++++++++++++++++++ ;+++++++++++++++++++++++++++++++++++++++++ folder = * FileTarget = folder + * FileResult = folder + * ;----------------------------------------- ;----------------------------------------- ; win = canny_gaussian(1.5, 8) ;define kernal win = fltarr(8,8) + 1 K = [0.05, 0.05] L = 100.0 ;read image finles++++++++++++++++++++++++ ;+++++++++++++++++++++++++++++++++++++++++ read_image_file ,FileTarget , data = DataTarget , samples = samples , lines = lines , bands = BandsTarget read_image_file ,FileResult , data = DataResult , samples = samples , lines = lines , bands = BandsResult ;----------------------------------------- ;----------------------------------------- ;calculate SSIM of each band++++++++++++++ ;+++++++++++++++++++++++++++++++++++++++++ SSIMTotal = 0 for Band = 0 , BandsTarget - 1dobegin; ;call the SobelRMSE procedure+++++++++++++ ;+++++++++++++++++++++++++++++++++++++++++ DataTargetBand DataResultBand = = DataTarget(*,*,Band) DataResult(*,*,Band) DataTargetBand = DataTargetBand[141:340,701:900] DataResultBand = DataResultBand[141:340,701:900] y = correlate(DataTargetBand,DataResultBand) calculate_SSIM, DataTargetBand, DataResultBand, DataMask, K, L, win, SSIMBand = SSIMBand SSIMTotal = SSIMTotal + SSIMBand ;----------------------------------------- ;----------------------------------------- print,SSIM of band + strtrim(string(Band+1),2)+ is : ,SSIMBand endfor ;----------------------------------------- ;----------------------------------------- end procalculate_SSIM, DataTarget, DataResult, DataMask, K, L, win, SSIMBand = SSIMBand C1 = (K[0]*L)^2 C2 = (K[1]*L)^2 win = win/total(win) Img1 = DataTarget Img2 = DataResult mu1 =convol(Img1,win,/EDGE_TRUNCATE) mu2 =convol(Img2,win,/EDGE_TRUNCATE)