PIL.Image.open与cv2.imread格式问题

常用图片读取方式：

PIL.Image.open

scipy.misc.imread

scipy.ndimage.imread

cv2.imread（速度最快）

matplotlib.image.imread

skimage

caffe.io.load_image

imagepath='test1.jpg'
im1=Image.open(imagepath)
im1=np.array(im1)#获得numpy对象,RGB
print(type(im1))
print(im1.shape)

#2 opencv 
im2=cv2.imread(imagepath)
print(type(im2))#numpy BGR
print(im2.shape)#[width,height,3]

#3 matplotlib 类似matlab的方式
im3 = mpimg.imread(imagepath)
print(type(im3))#np.array
print(im3.shape)

#4 skimge 
#caffe.io.load_iamge()也是调用的skimage实现的，返回的是0-1 float型数据
im4 = io.imread(imagepath)
print(type(im4))#np.array
print(im4.shape)
##########
# <class 'numpy.ndarray'> # (851, 1279, 3) 
# <class 'numpy.ndarray'> # (851, 1279, 3) 
# <class 'numpy.ndarray'> # (851, 1279, 3) 
# <class 'numpy.ndarray'> # (851, 1279, 3)