图像的重定义大小,图像的缩扩,图像的旋转:
1 from skimage importtransform,data2 importmatplotlib.pyplot as plt3 img =data.camera()4 print(img.shape)5 plt.subplot(221)6 plt.imshow(img)7 plt.subplot(222)8 plt.imshow(transform.resize(img,(64,64)))9 plt.subplot(223)10 plt.imshow(transform.rescale(img,0.2))11 plt.subplot(224)12 plt.imshow(transform.rotate(img,30,resize=True))13 plt.show()
产生高斯金字塔
1 importnumpy as np2 importmatplotlib.pyplot as plt3 from skimage importdata,transform4 image = data.astronaut() #载入宇航员图片
5 pyramid = transform.pyramid_gaussian(image, downscale=2) #产生高斯金字塔图像
6 #pyramid = transform.pyramid_laplacian(image, downscale=2)
7 #共生成了log(512)=9幅金字塔图像,加上原始图像共10幅,pyramid[0]-pyramid[1]
8 i = 1
9 for p inpyramid:10 plt.subplot(2,5,i)11 i+=1
12 #p[:,:,:]*=255
13 plt.title(p.shape)14 plt.imshow(p)15 plt.show()
gamma调整原理:I=Ig 如果gamma>1, 新图像比原图像暗。如果gamma<1,新图像比原图像亮
log对数调整I=log(I)
对比度是否偏低判断:exposure.is_low_contrast(img)
1 from skimage importdata, exposure, img_as_float2 importmatplotlib.pyplot as plt3 image =img_as_float(data.moon())4 gam1= exposure.adjust_gamma(image, 4) #调暗
5 gam2= exposure.adjust_gamma(image, 0.7) #调亮
6 gam3= exposure.adjust_log(image) #对数调整
7 plt.figure('adjust_gamma',figsize=(10,10))8 plt.subplot(141)9 plt.imshow(image)10 plt.subplot(142)11 plt.imshow(gam1)12 plt.subplot(143)13 plt.imshow(gam2,plt.cm.gray)14 plt.subplot(144)15 plt.imshow(gam3)16 plt.show() #原理:I=Ig
17 result=exposure.is_low_contrast(gam1)18 result
调整图片强度,不是很懂参数...
1 importnumpy as np2 from skimage importexposure3 image =data.moon()4 mat=exposure.rescale_intensity(image,out_range=(0,100))5 plt.subplot(121)6 plt.imshow(mat)7 print(image)8 print(mat)9 mat1=exposure.rescale_intensity(image, in_range=(0, 200))10 plt.subplot(122)11 plt.imshow(mat1)12 print(mat1.min())13 print(mat1)
绘制直方图
1 from skimage importdata2 importmatplotlib.pyplot as plt3 img=data.camera()4 plt.figure("hist")5 arr=img.flatten()6 n, bins, patches = plt.hist(arr, bins=256, normed=1,facecolor='red')7 plt.show()
彩色图片三通道直方图:
1 from skimage importdata2 importmatplotlib.pyplot as plt3 img=data.astronaut()4 ar=img[:,:,0].flatten()5 plt.hist(ar, bins=256, normed=1,facecolor='r',edgecolor='r',hold=1)6 ag=img[:,:,1].flatten()7 plt.hist(ag, bins=256, normed=1, facecolor='g',edgecolor='g',hold=1)8 ab=img[:,:,2].flatten()9 plt.hist(ab, bins=256, normed=1, facecolor='b',edgecolor='b')10 plt.show()
直方图均衡化exposure.equalize_hist(img)
对图像中像素个数多的灰度级进行展宽,而对图像中像素个数少的灰度进行压缩,从而扩展取值的动态范围,提高了对比度和灰度色调的变化,使图像更加清晰。
1 from skimage importdata,exposure2 importmatplotlib.pyplot as plt3 img=data.moon()4 plt.figure("hist",figsize=(8,8))5
6 arr=img.flatten()7 plt.subplot(221)8 plt.imshow(img,plt.cm.gray) #原始图像
9 plt.subplot(222)10 plt.hist(arr, bins=256, normed=1,edgecolor='None',facecolor='red') #原始图像直方图
11
12 img1=exposure.equalize_hist(img)13 arr1=img1.flatten()14 plt.subplot(223)15 plt.imshow(img1,plt.cm.gray) #均衡化图像
16 plt.subplot(224)17 arr1*=255
18 plt.hist(arr1, bins=256, normed=1,edgecolor='None',facecolor='red') #均衡化直方图
19
20 plt.show()
图像滤波:
平滑滤波,用来抑制噪声;微分算子,可以用来检测边缘和特征提取。
sobel、roberts、scharr、prewitt、canny算子
gabor、gaussian、median滤波
水平、垂直边缘检测
正负交叉边缘检测
1 from skimage importdata,filters,feature2 importmatplotlib.pyplot as plt3 from skimage.morphology importdisk4 img =data.camera()5 edges =filters.sobel(img)6 edges =filters.roberts(img)7 edges =filters.scharr(img)8 edges =filters.prewitt(img)9 edges = feature.canny(img,sigma=3)10 edges,filt_imag = filters.gabor(img, frequency=0.5)11 edges = filters.gaussian(img,sigma=5)12 edges = filters.median(img,disk(9))13 edges =filters.sobel_h(img)14 #水平边缘检测:sobel_h, prewitt_h, scharr_h
15 #垂直边缘检测: sobel_v, prewitt_v, scharr_v
16 edges =filters.roberts_neg_diag(img)17 edges =filters.roberts_pos_diag(img)18 plt.imshow(edges,plt.cm.gray)
图像阈值判断与分割的各种方法:
1 from skimage importdata,filters2 importmatplotlib.pyplot as plt3 image =data.camera()4 thresh =filters.threshold_otsu(image)5 thresh =filters.threshold_yen(image)6 thresh =filters.threshold_li(image)7 thresh =filters.threshold_isodata(image)8
9 dst =(image <= thresh)*1.0 #根据阈值进行分割
10 #dst =filters.threshold_adaptive(image, 31,'mean')
11 plt.subplot(121)12 plt.title('original image')13 plt.imshow(image,plt.cm.gray)14 plt.subplot(122)15 plt.title('binary image')16 plt.imshow(dst,plt.cm.gray)17 plt.show()
图形的绘制,与颜色。有各种各样的图形啊...
1 from skimage importdraw,data2 importmatplotlib.pyplot as plt3 img=data.chelsea()4 rr, cc=draw.ellipse(150, 150, 30, 80) #返回像素坐标5 draw.set_color(img,[rr,cc],[255,0,0])6 plt.imshow(img,plt.cm.gray)
图像的膨胀,腐蚀
1 from skimage importdata2 importskimage.morphology as sm3 importmatplotlib.pyplot as plt4 img=data.checkerboard()5 dst=sm.dilation(img,sm.square(5)) #用边长为15的正方形滤波器进行膨胀滤波
6 dst1=sm.erosion(img,sm.square(5)) #用边长为5的正方形滤波器进行膨胀滤波
7 plt.figure(figsize=(8,8))8 plt.subplot(131)9 plt.imshow(img,plt.cm.gray)10 plt.subplot(132)11 plt.imshow(dst,plt.cm.gray)12 plt.subplot(133)13 plt.imshow(dst1,plt.cm.gray)14 #找到像素值为1的点,将它的邻近像素点都设置成这个值。1值表示白,0值表示黑,因此膨胀操作可以扩大白色值范围,压缩黑色值范围。一般用来扩充边缘或填充小的孔洞
15 #将0值扩充到邻近像素。扩大黑色部分,减小白色部分。可用来提取骨干信息,去掉毛刺,去掉孤立的像素。
图像开运算,图像闭运算:
1 from skimage importio,color,data2 importskimage.morphology as sm3 importmatplotlib.pyplot as plt4 img=color.rgb2gray(data.camera())5 dst=sm.opening(img,sm.disk(9)) #用边长为9的圆形滤波器进行膨胀腐蚀滤波
6 dst1=sm.closing(img,sm.disk(9)) #用边长为5的圆形滤波器进行腐蚀膨胀滤波
7 plt.figure(figsize=(10,10))8 plt.subplot(131)9 plt.imshow(img,plt.cm.gray)10 plt.subplot(132)11 plt.imshow(dst,plt.cm.gray)12 plt.subplot(133)13 plt.imshow(dst1,plt.cm.gray)
白帽(white-tophat)。黑帽(black-tophat)。
1 from skimage importio,color2 importskimage.morphology as sm3 importmatplotlib.pyplot as plt4 img=color.rgb2gray(data.camera())5 dst=sm.white_tophat(img,sm.square(21)) #将原图像减去它的开运算值,返回比结构化元素小的白点
6 dst1=sm.black_tophat(img,sm.square(21)) #将原图像减去它的闭运算值,返回比结构化元素小的黑点,且将这些黑点反色。
7 plt.figure('morphology',figsize=(10,10))8 plt.subplot(131)9 plt.imshow(img,plt.cm.gray)10 plt.subplot(132)11 plt.imshow(dst,plt.cm.gray)12 plt.subplot(133)13 plt.imshow(dst1,plt.cm.gray)