基于Python的PIL库学习详解
摘要
对于图像识别，大量的工作在于图像的处理，处理效果好，那么才能很好地识别，因此，良好的图像处理是识别的基础。在Python中，有一个优秀的图像处理框架，就是PIL库，本博文会分模块，介绍PIL库中的各种方法，并列举相关例子。
参考：http://pillow-cn.readthedocs.io/zh_CN/latest/reference/index.html
网站上列举了PIL库中所有的模块和方法，但是没有相关的例子，博文中会尽量给出相关的例子和进行简单的讲解。
基于的环境：Win10，Python2.7，PIL1.1.7。
Image模块
开篇的例子
首先，给出Image模块中的一个简单的例子。例子实现的功能是：读取图片，并进行45°旋转，然后进行可视化。
#-*-coding:utf-8-*-
#Image模块开篇例子
fromPILimportImage
im=Image.open('test.bmp')#读取图片
im.rotate(45).show()#将图片旋转，并用系统自带的图片工具显示图片
创建缩略图
#-*-coding:utf-8-*-
#PIL中创建缩略图(createthumbnails)
fromPILimportImage
importglob,os
size=128,128
forinfileinglob.glob("*.jpg"):#glob的作用是文件搜索，返回的是一个列表
file,ext=os.path.splitext(infile)#将文件的文件名和拓展名分开，用于之后的保存重命名
im=Image.open(infile)
im.thumbnail(size,Image.ANTIALIAS)#等比例缩放
im.save(file+".thumbnail","JPEG")
#im.show()#显示缩略图
#printim.size,im.mode
缩略图不能直接双击打开，而可以使用PIL.image的open读取，然后使用show()方法进行显示。
图像处理
PIL.image.alpha_composite(im1,im2) PIL.image.blend(im1,im2,alpha) PIL.Image.composite(im1,im2,mask)
这三个方法都属于图片的合成或者融合。都要求im1和im2的mode和size要一致，alpha代表图片占比的意思，而mask是mode可以为”1”,”L”或者”RGBA”的size和im1、im2一致的。
#coding:utf-8-*-
fromPILimportImage
#图片合成
#PIL的alpha_composite(im1,im2)图像通道融合
#im2要和im1的size和mode一致，且要求格式为RGBA
im1=Image.open("test.png")
im2=Image.open("test2.png")
newim1=Image.alpha_composite(im1,im2)#将im2合成到im1中，如果其一是透明的，
#才能看到结果，不然最终结果只会显示出im2
newim1.show()
#print(im1.mode)
#-----------------------------------------
#image.blend(im1,im2,alpha)
#alpha为透明度
newim2=Image.blend(im1,im2,0.5)
newim2.show()
#-----------------------------------------
mask=Image.open("mask.png")
newim3=Image.composite(im2,im1,mask)
newim3.show()
PIL.image.eval(image,*args)
程序：
#-*-coding:utf-8-*-
fromPILimportImage
im=Image.open("test.png")
imnew=Image.eval(im,lambdai:i*2)#将原图片的像素点，都乘2，返回的是一个Image对象
#printimnew.mode
imnew.show()
im.show()
创建图像
(1)PIL.image.new(mode,size,color=0)
使用模式和大小，创建一个新的图像。其中，mode可以是”L”,”RGB”,”RGBA”;而size则是一个tuple(元组),color应该和mode相对应。
下面例子，分别创建”L”、”RGB”和”RGBA”的图片。
#-*-coding:utf-8-*-
fromPILimportImage
#创建图像
#创建一个灰度图像
newL=Image.new("L",(28,28),255)
newL.show()
#创建一个RGb图像
newrgb=Image.new("RGB",(28,28),(20,200,45))
newrgb.show()
newrgba=Image.new("RGBA",(28,28),(20,200,45,255))
newrgba.show()
print"Thefristimage:",newL.size,newL.mode
print"Thesecondimage:",newrgb.size,newrgb.mode
print"Thethirdimage:",newrgba.size,newrgba.mode
(2)以其他形式创建图像
a.以数组的形式创建图像，PIL.image.fromarray(obj,mode=None)
obj-图像的数组，类型可以是numpy.array()
mode-如果不给出，会自动判断
本人觉得这个功能还是挺实用的，可以将一个数组（具体一点就是像素数组）转换为图像，从图像的本质去处理图像。
下面一段程序，就是用fromarray()函数实现图像的灰度化（使用了两种方法）。
#-*-coding:utf-8-*-
fromPILimportImage
importnumpyasnp
a=Image.open("fromimg.png")
a.show()
b=a.resize((28,28))
datab=list(b.getdata())
#printtype(datab)
obj1=[]
obj2=[]
foriinrange(len(datab)):
obj1.append([sum(datab[i])/3])#灰度化方法1：RGB三个分量的均值
obj2.append([0.3*datab[i][0]+0.59*datab[i][1]+0.11*datab[i][2]])
#灰度化方法2：根据亮度与RGB三个分量的对应关系：Y=0.3*R+0.59*G+0.11*B
obj1=np.array(obj1).reshape((28,28))
obj2=np.array(obj2).reshape((28,28))
printobj1
printobj2
arrayimg1=Image.fromarray(obj1)
arrayimg2=Image.fromarray(obj2)
arrayimg1.show()
arrayimg2.show()
显然，两种方法都能成功灰度化。
还有：
PIL.Image.frombytes(mode,size,data,decoder_name='raw',*args) PIL.Image.fromstring(*args,**kw) PIL.Image.frombuffer(mode,size,data,decoder_name='raw',*args)
感觉不常用，没有仔细研究。
Image模块下的Image类
下面的Image是一个图像对象，而不是模块！
(1)Image.convert(mode=None,matrix=None,dither=None,palette=0,color=256)
该方法，同样可以实现上面的灰度化处理。
#-*-coding:utf-8-*-
fromPILimportImage
img=Image.open("test.png")
#灰度化：将RGB/RGBA->L
img=img.convert("L")
img.show()
(2)Image.copy()
将读取的图片复制一份。
#-*-coding:utf-8-*-
fromPILimportImage
img=Image.open("test.png")
#灰度化：将RGB/RGBA->L
img=img.convert("L")
#img.show()
#------copy()----------
img1=img.copy()
img1.show()
将灰度化的图片复制一份，因此该程序的运行结果和之前的一致。
(3)Image.filter(filter)
该函数是用于图像滤波的，PIL中自带了很多的滤波器，就是括号中的filter的参数。filter应该是一个ImaageFilter模块下的对象。这里把ImageFilter模块讲了。其实，该模块就是提供滤波器。自带的滤波器有：
使用中值滤波：
#-*-coding:utf-8-*-
fromPILimportImage
fromPILimportImageFilter
#BLUR-模糊处理
#CONTOUR-轮廓处理
#DETAIL-增强
#EDGE_ENHANCE-将图像的边缘描绘得更清楚
#EDGE_ENHANCE_NORE-程度比EDGE_ENHANCE更强
#EMBOSS-产生浮雕效果
#SMOOTH-效果与EDGE_ENHANCE相反，将轮廓柔和
#SMOOTH_MORE-更柔和
#SHARPEN-效果有点像DETAIL
testimg=Image.open("filter1.png")
testimg.show()
filterimg=testimg.filter(ImageFilter.MedianFilter)
filterimg.show()
(4)使用各种方法/函数获取图片的基本信息
Image.getbands()
Image.geebbox()
Image.getcolors(maxcolor=256)
Image.getdata(band=None)（一般和list()结合使用）
Image.getextrema()
Image.getpixel((x,y))
Image.histogram(mask=None,extrema=None)
#-*-coding:utf-8-*-
fromPILimportImage
img1=Image.open("test.png")
img1.show()
#getbands()-显示该图像的所有通道，返回一个tuple
bands=img1.getbands()
printbands
#getbbox()-返回一个像素坐标，4个元素的tuple
bboxs=img1.getbbox()
printbboxs
#getcolors()-返回像素信息，是一个含有元素的列表[(该种像素的数量，(该种像素)),(...),...]
colors=img1.getcolors()
printcolors
#getdata()-返回图片所有的像素值，要使用list()才能显示出具体数值
#data=list(img1.getdata())
#printdata
#getextrema()-获取图像中每个通道的像素最小和最大值,是一个tuple类型
extremas=img1.getextrema()
printextremas
#getpixel()-获取该坐标
pixels=img1.getpixel((87,180))
printpixels
#histogram()-返回图片的像素直方图
print(img1.histogram())
运行结果:
('R','G','B','A')
(0,0,338,238)
[(73463,(255,255,255,255)),(32,(252,249,252,255)),(1,(255,189,143,255)),(12,(255,199,160,255)),(22,(247,239,247,255)),(3,(255,242,246,255)),(9,(238,221,238,255)),(9,(235,215,235,255)),(5,(232,209,232,255)),(1,(255,228,209,255)),(2,(255,210,225,255)),(1,(255,202,201,255)),(3,(255,158,92,255)),(22,(218,181,218,255)),(1,(217,181,218,255)),(2,(255,232,217,255)),(16,(255,195,153,255)),(22,(212,169,212,255)),(3,(211,169,212,255)),(1,(204,153,204,255)),(1,(255,229,238,255)),(53,(255,131,46,255)),(9,(255,203,167,255)),(1,(255,157,90,255)),(3,(186,119,187,255)),(2,(255,217,229,255)),(6,(183,113,184,255)),(1,(255,212,227,255)),(14,(214,175,215,255)),(2,(255,182,131,255)),(12,(166,79,167,255)),(2,(255,180,127,255)),(4309,(255,127,39,255)),(737,(163,73,164,255)),(4,(255,252,253,255)),(3,(255,232,216,255)),(9,(255,250,233,255)),(1,(255,245,248,255)),(34,(255,239,228,255)),(3,(255,142,64,255)),(1,(255,162,98,255)),(19,(255,247,241,255)),(7,(255,223,201,255)),(2,(255,133,49,255)),(16,(255,221,232,255)),(58,(255,235,221,255)),(1,(255,225,204,255)),(2,(255,219,194,255)),(21,(255,175,120,255)),(6,(255,182,206,255)),(37,(255,243,235,255)),(3,(255,179,127,255)),(6,(255,207,223,255)),(3,(255,232,240,255)),(1,(255,134,51,255)),(2,(255,222,233,255)),(2,(255,218,192,255)),(1,(255,186,186,255)),(1,(255,163,99,255)),(1,(255,207,173,255)),(8,(255,151,80,255)),(1,(255,184,201,255)),(19,(255,211,180,255)),(1,(255,143,65,255)),(9,(255,233,158,255)),(18,(255,215,187,255)),(1,(255,185,136,255)),(7,(255,227,237,255)),(22,(255,163,100,255)),(1,(255,221,198,255)),(5,(255,184,208,255)),(10,(255,195,215,255)),(5,(255,239,182,255)),(1,(255,197,157,255)),(1,(255,154,85,255)),(1,(255,136,55,255)),(8,(255,240,190,255)),(14,(255,216,229,255)),(3,(255,179,204,255)),(1,(255,143,67,255)),(1,(255,196,155,255)),(19,(255,249,227,255)),(2,(255,211,181,255)),(10,(255,230,142,255)),(4,(255,187,140,255)),(195,(255,201,14,255)),(2,(255,129,42,255)),(1,(255,131,47,255)),(12,(255,231,214,255)),(1,(255,181,151,255)),(8,(249,244,249,255)),(13,(246,238,246,255)),(44,(244,234,244,255)),(1,(243,232,244,255)),(7,(240,226,241,255)),(25,(255,167,107,255)),(24,(255,215,229,255)),(22,(230,206,230,255)),(6,(229,204,229,255)),(3,(255,130,45,255)),(11,(227,200,228,255)),(4,(226,198,226,255)),(3,(255,127,40,255)),(5,(223,192,223,255)),(9,(220,186,221,255)),(172,(255,174,201,255)),(16,(255,231,239,255)),(1,(255,171,113,255)),(33,(209,164,209,255)),(1,(255,192,213,255)),(6,(255,247,250,255)),(2,(255,136,54,255)),(9,(255,253,247,255)),(1,(255,171,114,255)),(2,(255,147,73,255)),(5,(255,181,130,255)),(7,(189,124,190,255)),(1,(255,199,161,255)),(13,(255,183,134,255)),(3,(255,152,82,255)),(2,(255,156,88,255)),(32,(255,143,66,255)),(5,(178,102,178,255)),(6,(175,96,176,255)),(8,(255,129,43,255)),(4,(172,90,173,255)),(1,(255,168,109,255)),(1,(255,153,83,255)),(1,(255,174,118,255)),(1,(255,172,115,255)),(1,(255,148,75,255)),(8,(255,244,248,255)),(1,(255,130,43,255)),(5,(255,205,222,255)),(1,(255,210,177,255)),(1,(255,170,110,255)),(1,(255,157,89,255)),(1,(255,197,134,255)),(13,(255,155,86,255)),(3,(255,137,56,255)),(2,(255,138,57,255)),(11,(255,227,208,255)),(1,(255,190,145,255)),(2,(255,155,87,255)),(1,(169,84,170,255)),(4,(255,202,220,255)),(6,(255,139,59,255)),(1,(255,128,42,255)),(1,(255,158,91,255)),(1,(255,198,158,255)),(5,(255,130,44,255)),(1,(255,202,165,255)),(1,(255,187,154,255)),(1,(255,132,48,255)),(1,(255,154,84,255)),(1,(255,235,241,255)),(7,(255,135,53,255)),(62,(255,159,93,255)),(2,(255,177,124,255)),(4,(255,187,210,255)),(11,(255,251,248,255)),(1,(255,229,211,255)),(1,(255,208,176,255)),(1,(255,133,50,255)),(2,(255,219,231,255)),(2,(255,141,63,255)),(2,(255,146,71,255)),(1,(255,160,95,255)),(2,(255,184,135,255)),(1,(255,208,175,255)),(1,(255,139,61,255)),(1,(255,189,211,255)),(2,(255,145,69,255)),(263,(255,191,147,255)),(4,(255,187,141,255)),(3,(255,250,252,255)),(1,(255,147,72,255)),(5,(255,177,203,255)),(1,(255,169,109,255)),(62,(255,207,174,255))]
((163,255),(73,255),(14,255),(255,255))
(255,127,39,255)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,737,0,0,12,0,0,1,0,0,4,0,0,6,0,0,5,0,0,0,0,6,0,0,3,0,0,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,33,0,3,22,0,14,0,0,1,22,0,9,0,0,5,0,0,4,11,0,6,22,0,5,0,0,9,0,0,9,0,7,0,0,1,44,0,13,22,0,8,0,0,32,0,0,79360,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,737,0,0,0,0,0,12,0,0,0,0,1,0,0,0,0,0,4,0,0,0,0,0,6,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,3,0,0,0,0,7,0,0,4312,1,10,9,54,1,3,1,7,3,3,2,7,0,2,3,34,0,2,2,3,1,0,0,8,3,2,2,15,2,2,4,62,1,0,1,23,33,0,0,25,1,26,1,2,1,0,173,35,0,7,0,6,2,29,8,13,8,1,10,13,0,2,1,263,6,0,0,26,1,2,5,13,11,195,6,9,6,5,22,69,2,5,3,21,1,0,0,51,14,2,2,4,0,26,2,7,0,1,7,18,1,2,10,28,9,9,44,59,0,0,13,61,8,0,3,37,16,1,0,25,0,51,12,11,4,9,0,73463,0,0,0,0,0,0,0,0,0,0,0,0,0,0,195,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4309,3,0,3,9,5,3,53,1,1,2,1,1,0,7,2,1,3,2,0,6,0,1,0,2,3,1,32,1,0,2,0,2,1,2,0,1,0,0,0,0,8,0,3,1,1,1,13,2,2,1,1,1,3,62,0,1,0,0,1,1,22,0,0,0,0,0,0,25,0,2,1,0,0,1,1,1,0,0,1,0,21,0,0,0,2,0,0,5,0,0,5,2,0,0,14,2,1,0,0,0,4,4,10,1,0,1,0,263,0,0,0,1,0,16,1,1,0,1,10,0,12,1,0,0,737,1,0,21,0,0,1,0,0,5,62,1,7,1,5,0,19,2,5,0,6,0,1,21,0,0,15,0,2,0,2,0,0,0,1,0,0,181,0,5,5,0,6,0,16,34,4,2,25,1,12,24,3,2,23,0,4,67,5,11,0,2,4,20,45,46,22,2,21,11,0,46,0,7,10,16,3,27,0,0,45,0,16,31,20,8,6,0,35,4,0,73463,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,80444]
(5)图像粘贴操作(paste)
Image.paste(im,box=None,maske=None)
使用im粘贴到原图片中。注意：两个图片的mode和size要求一致，不一致可以使用convert()和resize()进行调整。
#-*-coding:utf-8-*-
fromPILimportImage
rawimg=Image.open("qqtou.png")
printrawimg.size
im=Image.open("number.png")
printim.size
#rawimg的size和im的size要相同，不然不能匹配
#paste(用来粘贴的图片，(位置坐标))，可以通过设置位置坐标来确定粘贴图片的位置
#该方法没有返回值，直接作用于原图片
rawimg.paste(im,(75,-90))
rawimg.show()
(6)各种put操作
Image.putalpha(alpha)-添加多一层alpha层，没看出具体效果
Image.putdata(data,scale=1.0,offset=0.0)-添加一个像素序列到原图像。
#-*-coding:utf-8-*-
fromPILimportImage
img=Image.open("qqtou.png")
img=img.convert("L")
img.show()
imgdata=list(img.getdata())
printimgdata
addlist=[]
foriinrange(len(imgdata)):
ifimgdata[i]>250:
addlist.append(imgdata[i]-100)
else:
addlist.append(imgdata[i])
#putdata-将一个序列添加进原图像，没有返回值，直接作用在原图像中
img.putdata(addlist)
img.show()
显然，原始图像（左图）已经发生了改变。
以上就是本文的全部内容，希望对大家的学习有所帮助，也希望大家多多支持毛票票。