图象增强-对比度.ppt

,单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,-,*,第三章 图象增强,对比度增强,图象平滑,图象锐化,同态滤波,伪彩色与假彩色处理,1,-,2,-,目的：,采用一系列技术去改善图象的视觉效果，或将图象转换成一种更适合于人或机器进行分析处理的形式。图象增强并不以图象保真为准则，而是有选择地突出某些对人或机器分析有意义的信息，抑制无用信息，提高图象的使用价值。,3,-,方法：,空间域处理,全局运算：,在整个图象空间域进行,对整幅图象进行相同的处理，如旋转、放大、缩小等,。,局部运算：,在与象素有关的空间域进行，根据以处理象素中心的邻域多个象素灰度值计算象素值，如均值滤波、中值滤波等。,点运算：,对图象作逐点运算，输出只依赖于对应点的灰度值，如对比度操作或对比度拉伸。,频域处理,在图象的,Fourier,变换域上进行处理。,4,-,灰度变换法,线性变换,对数变换,指数变换,直方图调整法,直方图均衡化,直方图匹配,3.1,对比度增强,5,-,(,一)线性灰度变换,当图象成象时曝光不足或过度,或由于成象设备的非线性和图象记录设备动态范围太窄等因素。都会产生对比度不足的弊病，使图象中的细节分辨不清。这时可将,灰度范围线性扩展,。,设,f(x,y),灰度范围为,a,b，g(x,y),灰度范围为,c,d,灰度变换法,6,-,7,-,0,f(x,y),g(x,y),a,b,c,d,8,-,灰度变换法,(二)分段线性灰度变换,将感兴趣的灰度范围线性扩展，相对抑制不感兴趣的灰度区域。,设,f(x,y),灰度范围为,0,M,f,，g(x,y),灰度范围为,0,M,g,9,-,分段线性灰度变换,10,-,分段线性灰度变换,11,-,%,Matlab,代码,I=imread(lina.bmp);,%,读图象,M,N=size(I);,%,获取图象大小,R=zeros(M,N);,%,初始化结果图象,for i=1:M,%,行,for j=1:N,%,列,g=I(i,j);,%,当前象素灰度,if g=a)&(gb),R(i,j)=(d-c)*(g-a)/(b-a)+c;,else,R(i,j)=(Mg-d)*(g-b)/(Mf-b)+d;,end;,end;,end;,12,-,灰度变换法,(,三)非线性灰度变换,（1）对数变换,低灰度区扩展，高灰度区压缩。,（2）,指数变换,高灰度区扩展，低灰度区压缩。,13,-,a,b,c,是按需要可以调整的参数,。,对数变换,14,-,对数变换,15,-,a,b,c,是按需要可以调整的参数。,指数变换,16,-,指数变换,17,-,原始图象,灰度变换示例,18,-,非线性灰度变换,对数效应,低灰度区扩展，高灰度区压缩,19,-,非线性灰度变换,指数效应,低灰度区压缩，高灰度区扩展,20,-,分段线性化,出现假轮廓,21,-,灰度倒置,底片效果,22,-,阈值化,阈值128,23,-,阈值化,阈值180,24,-,阈值化,阈值66,25,-,Matlab,中灰度调整,函数,%,例1 灰度图象,I=imread(pout.tif);,J=,imadjust,(I,0.3 0.7,);,imshow(I),figure,imhist(I);figure,imshow(J),figure,imhist(J);,imadjust,26,-,27,-,28,-,%,例2 彩色图象,RGB1=imread(flowers.tif);,RGB2=imadjust(RGB1,.2.3 0;.6.7 1,);,imshow(RGB1),figure,imshow(RGB2),29,-,30,-,（,一）直,方图均衡化,Histogram Equalization,直,方图,:表示数字图象中的每一灰度级与其出现的频率(该灰度级的象素数目)间的统计关系,用横坐标表示灰度级,纵坐标表示频数(也可用概率表示)。,直方图调整法,31,-,直,方图,32,-,I=imread(rice.tif);,M,N=size(I);,s=256;%,灰度级,H=zeros(1,s);,for i=1:M,for j=1:N,g=I(i,j);,H(1,g+1)=H(1,g+1)+1;,end;,end;,计算直方图,33,-,灰度直,方图,34,-,彩色直,方图,35,-,直方图均衡化是将原图象的直方图通过变换函数修正为均匀的直方图，然后按均衡直方图修正原图象。,图象均衡化处理后，图象的直方图是平直的，即各灰度级具有相同的出现频数，那么由于灰度级具有均匀的概率分布，图象看起来就更清晰了。,直,方图均衡化,36,-,直,方图均衡化,首先假定,连续灰度级,的情况，推导直方图均衡化变换公式，令,r,代表灰度级，,P(r),为概率密度函数。,r,值已归一化，最大灰度值为1。,灰度级数为,0,1,L-1,共,L,级。,37,-,连续灰度的直,方图,非均匀分布,38,-,连续灰度的直,方图,均匀分布,39,-,直,方图均衡化,目标,直,方图均衡化,40,-,直,方图均衡化,要找到一种变换,S,=,T,(,r,),使直方图变平直，为使变换后的灰度仍保持从黑到白的单一变化顺序，且变换范围与原先一致，以避免整体变亮或变暗。必须规定：,（1）在0,r,1,中，,T,(,r,),是单调递增函数，且0,T,(,r,)1；,（2）,反变换,r,=,T,-1,(,s,),T,-1,(,s,),也为单调递增函数，0,s,1。,41,-,r,j,r,j,+,r,s,j,s,j,+,s,直,方图均衡化,变换公式推导图示,42,-,直,方图均衡化,考虑到灰度变换不影响象素的位置分布，也不会增减象素数目。所以有,43,-,直,方图均衡化,应用到离散灰度级，设一幅图象的象素总数为,n,，,分,L,个灰度级。,n,k,:,第,k,个灰度级出现的频数。,第,k,个灰度级出现的概率,P,(,r,k,)=,n,k,/,n,其中0,r,k,1，,k,=0,1,2,.,L,-1,形式为：,44,-,例：设图象有64*64=4096个象素，有8个灰度级，灰度分布如表所示。进行,直,方图均衡化,。,r,k,r,0,=0,r,1,=1/7,r,2,=2/7,r,3,=3/7,r,4,=4/7,r,5,=5/7,r,6,=6/7,r,7,=1,n,k,790,1023,850,656,329,245,122,81,p,(r,k,),0.19,0.25,0.21,0.16,0.08,0.06,0.03,0.02,45,-,1.由（3-2）式计算,s,k,r,k,r,0,=0,r,1,=1/7,r,2,=2/7,r,3,=3/7,r,4,=4/7,r,5,=5/7,r,6,=6/7,r,7,=1,n,k,790,1023,850,656,329,245,122,81,p(r,k,),0.19,0.25,0.21,0.16,0.08,0.06,0.03,0.02,s,k,计算,0.19,0.44,0.65,0.81,0.89,0.95,0.98,1.00,46,-,r,k,r,0,=0,r,1,=1/7,r,2,=2/7,r,3,=3/7,r,4,=4/7,r,5,=5/7,r,6,=6/7,r,7,=1,n,k,790,1023,850,656,329,245,122,81,p,(r,k,),0.19,0.25,0.21,0.16,0.08,0.06,0.03,0.02,s,k,计算,0.19,0.44,0.65,0.81,0.89,0.95,0.98,1.00,s,k,舍入,1/7,3/7,5/7,6/7,6/7,1,1,1,2.把计算的,s,k,就近安排到8个灰度级中,47,-,r,k,r,0,=0,r,1,=1/7,r,2,=2/7,r,3,=3/7,r,4,=4/7,r,5,=5/7,r,6,=6/7,r,7,=1,n,k,790,1023,850,656,329,245,122,81,p(r,k,),0.19,0.25,0.21,0.16,0.08,0.06,0.03,0.02,s,k,计算,0.19,0.44,0.65,0.81,0.89,0.95,0.98,1.00,s,k,舍入,1/7,3/7,5/7,6/7,6/7,1,1,1,s,k,s,0,s,1,s,2,s,3,s,4,n,sk,790,1023,850,985,448,p(s,k,),0.19,0.25,0.21,0.24,0.11,3.重新命名,s,k，,归并相同灰度级的象素数,48,-,直,方图均衡化,均衡化前后,直,方图比较,49,-,直,方图均衡化,直方图均衡化实质上是,减少图象的灰度级以换取对比度的加大,。在均衡过程中，原来的直方图上频数较小的灰度级被归入很少几个或一个灰度级内，故得不到增强。若这些灰度级所构成的图象细节比较重要，则需采用局部区域直方图均衡。,50,-,直,方图均衡化,实例1,51,-,直,方图均衡化,实例2,52,-,Matlab,中有关直,方图的函数,1、显示直方图,imhist(),I=imread(pout.tif);,imshow(I);,figure;,imhist,(I);,53,-,54,-,%例1,J=,histeq,(I);,imshow(I);title(,原始图象);,figure,imshow(J);,title(,直方图均衡化后图象);,2、直方图均衡化,histeq(),55,-,56,-,57,-,%例2,I=imread(tire.tif);,J=,histeq,(I);,imshow(I);,figure,imshow(J),58,-,59,-,60,-,Terms,Image enhancement:,图象增强,Image quality:,图象质量,Algorithm:,算法,Globe operation:,全局运算,Local operation:,局部运算,Point operation:,点运算,Spatial:,空间的,Spatial domain:,空间域,Spatial coordinate:,空间坐标,61,-,Terms,Linear:,线性,Nonlinear:,非线性,Frequency:,频率,Frequency variable:,频率变量,Frequency domain:,频域,Fourier transform:,傅立叶变换,One-dimensional Fourier transform:,一维傅立叶变换,62,-,Terms,Two-dimensional Fourier transform:,二维傅立叶变换,Discrete Fourier transform(DFT):,离散傅立叶变换,Fast Fourier transform(FFT):,快速傅立叶变换,Inverse Fourier transform:,傅立叶反变换,Contrast enhancement:,对比度增强,Contrast stretching:,对比度扩展,63,-,Terms,Gray-scale transformation(GST):,灰度变换,Logarithm transformation:,对数变换,Exponential transformation:,指数变换,Threshold:,阈值,Thresholding:,二值化、门限化,False contour:,假轮廓,64,-,Terms,Histogram:,直方图,Multivariable histogram:,多变量直方图,Histogram modification:,直方图调整、直方图修改,Histogram equalization:,直方图均衡化,Histogram specification:,直方图规定化,Histogram matching:,直方图匹配,65,-,Terms,Histogram thresholing:,直方图门限化,Probability density function(PDF):,概率密度函数,Cumulative distribution function(CDF):,累积分布函数,Slope:,斜率,Normalized:,归一化,Inverse function:,反函数,66,-,Terms,Calculus:,微积分,Derivative:,导数,Integral:,积分,Monotonic function:,单调函数,Infinite:,无穷大,Infinitesimal:,无穷小,Equation:,方程,67,-,