Python中，如果你遇到了PIL图像保存有白边，CV2.polyline，fillpoly，参数问题，图像保存颜色发生异常这几个问题，这篇文章就能够解决你的疑惑。

第一个问题，plt图像保存有白边

首先，plt图像保存有白边，设置savefig里的参数和plt.tight_layout都无法真正的去除,plt适合画图表，有坐标值的这种。

即使按照这种方案https://blog.csdn.net/jifaley/article/details/79687000，也是没用的，为什么呢？首先，plt是封装了matlab中的plot包，发现matlab中也是有这个问题，matlab解决方案http://blog.sina.com.cn/s/blog_66d4b4620101fvph.html，因为plt.save()时是将整个窗口保存下来的

在python的引用格式：import matplotlib.pyplot as plt

#接下来在试了很多方法后无果

plt.axis('off')

subplots_adjust(left=None, bottom=None, right=None, top=None, wspace=None, hspace=None)

#这样都是无法真正去除掉的

改用CV2就可以完美的解决这个问题，试一下后，不错~，因为opencv展示时，是按照图片的大小来展示的

im = cv2.imread('demo.jpg')

cv2.imshow('image',im)

cv2.imwrite('aa.jpg',img,[int(cv2.IMWRITE_JPEG_QUALITY),95])

cv2.waitKey(0)

cv2.destroyAllWindows()

第二个问题，CV2.polyline，fillpoly的参数

python中cv2.polyline和cv2.fillPoly参数如下：

image.png

很不错的参考链接关于poly的，https://blog.csdn.net/lkiller_hust/article/details/52949020

polyline中

第一个参数是img图像，表示你要在哪张图像上画线

pts表示的是点对，形式如下：

[[[334 33]]

[[ 44 255]]

[[444 222]]

[[ 33 333]]]

isClosed，布尔型，True表示的是线段闭合，False表示的是仅保留线段

color，线段颜色，格式是(R,G,B)值

thickness, 数值型，厚度，默认值为1，如果对封闭图形，正方形，三角形等传入-1,则会填充整个图形。

lineType，线型

shift,默认为0，

COCO_train2014_000000325939.jpg

举一个搞笑唧唧的例子

import cv2

import glob

import numpy as np

from PIL import Image

import matplotlib

import random

#im = np.array(Image.open('COCO_train2014_000000325939.jpg'))

im = cv2.imread('COCO_train2014_000000325939.jpg')

#w,h = im.size

#img = np.array(Image.new('RGB', (w, h), 'black'))

anns = [[184.83,393.26,186.02,387.29,186.62,386.1,195.57,381.92,197.36,380.73,200.93,377.75,205.71,374.17,

209.29,369.39,213.46,365.22,219.43,362.24,224.8,359.25,236.73,355.67,239.11,355.67,248.66,359.25,

252.83,362.83,257.01,365.82,261.78,367.6,274.31,370.59,288.62,375.96,301.15,378.94,310.7,382.52,

319.64,386.1,325.01,389.08,335.75,395.64,347.68,400.41,365.58,403.99,369.15,406.38,378.1,414.13,

388.24,420.1,400.17,427.26,409.72,433.82,410.91,435.61,389.44,427.26,376.91,426.66,368.56,426.66,

376.91,433.22,381.08,437.4,386.45,442.17,388.24,445.75,369.15,445.75,360.8,442.77,348.87,441.57,

333.96,442.17,322.63,442.77,310.7,443.96,299.96,446.94,291.61,451.71,290.41,454.1,289.22,463.05,289.22,473.79,

285.64,481.54,283.85,494.66,283.25,498.84,273.11,503.61,262.38,501.82,260.59,501.23,254.03,499.44,245.08,500.63,

230.76,497.65,235.53,495.86,245.08,495.26,252.83,494.07,259.39,491.68,262.38,487.51,262.97,483.93,265.36,479.15,

267.75,470.21,268.94,462.45,270.13,455.89,264.76,449.33,259.39,447.54,254.62,445.15,245.67,440.98,230.16,430.24,

224.2,426.66,218.83,422.48,215.25,413.54,214.06,409.96,213.46,406.38,213.46,398.03,212.87,393.26,207.5,390.87,

206.9,390.87,199.74,390.27,196.16,393.26,190.79,395.64,187.81,395.64],[334,33,44,255,444,222,33,333]

]

def showskt(img,anns):

#c = (np.random.random((1, 3))*0.6+0.4).tolist()[0]

if len(anns) == 0:

return 0

polygons = []

color = []

for ann in anns:

c = np.random.randint(0, 256, 3)

print(c)

#c = (np.random.random((1, 3)) * 0.6 + 0.4).tolist()[0]

pts = np.array(ann).reshape((int(len(ann) / 2), 2))

pts = pts.reshape((-1, 1, 2))

print(pts)

#img = cv2.fillPoly(im,np.int32([pts]),c,cv2.LINE_AA)

img = cv2.polylines(img, np.int32([pts]), True, c, -1, cv2.LINE_AA)

cv2.imshow('image',img)

cv2.imwrite('aa.jpg',img,[int(cv2.IMWRITE_JPEG_QUALITY),95])

cv2.waitKey(0)

cv2.destroyAllWindows()

showskt(im,anns)

看看效果吧，没有白边，线也画出来了，同样，分割的话，就去掉我注释掉的#

img = cv2.fillPoly(im,np.int32([pts]),c,cv2.LINE_AA)就好了

image.png

混用CV和PIL图像读取，图像显示，发生保存颜色发生异常

这是由于plt和cv2图像的通道顺序是不一样的，所以交换第一通道和第三通道就可以了

import numpy as np

import cv2

import matplotlib.pyplot as plt

img = cv2.imread('demo.jpg')

(r, g, b)=cv2.split(img)

img=cv2.merge([b,g,r])

plt.imshow(img)

plt.show()