python中 OpenCV和Pillow处理图像操作及时间对比

import cv2
from PIL import Image

img_path = 'images/test_demo.png'

cv_img = cv2.imread(img_path)
height, width, channel = cv_img.shape

pillow_img = Image.open(img_path)
width, height = pillow_img.size

import cv2
from PIL import Image
import numpy as np

png_img_path = 'images/test_demo.png'
jpg_img_path = 'images/test_demo.jpg'

# 由jupyter notebook中魔法命令：%%timeit测得
# 169 ms ± 1.68 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
cv_img = cv2.imread(png_img_path)    

# 52.9 ms ± 541 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
cv_img = cv2.imread(jpg_img_path)

# 300 ms ± 8.45 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
pillow_img = Image.open(png_img_path)
pillow_img = np.array(pillow_img)

# 47.4 ms ± 1.87 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
pillow_img = Image.open(jpg_img_path)
pillow_img = np.array(pillow_img)

save_png_path = 'output/result.png'
save_jpg_path = 'output/result.jpg'

cv_img = cv2.imread(png_img_path)
pillow_img = Image.open(png_img_path)

# 346 ms ± 11.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
cv2.imwrite(save_png_path, cv_img)

# 158 ms ± 4.03 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
cv2.imwrite(save_jpg_path, cv_img)

# 2.81 s ± 38.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
pillow_img.save(save_png_path)

# 51.3 ms ± 1.72 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
t = pillow_img.convert('RGB') 
t.save(save_jpg_path)

png_img_path = 'images/test_demo.png'

resize_shape = (2048, 2048)
cv_img = cv2.imread(png_img_path)
pillow_img = Image.open(png_img_path)

# 6.93 ms ± 173 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
cv2.resize(cv_img, resize_shape, interpolation=cv2.INTER_CUBIC)

# 151 ms ± 2.21 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
pillow_img.resize(resize_shape, resample=Image.Resampling.BICUBIC)

angle = 38

# 23.6 ms ± 732 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
h, w = cv_img.shape[:2]
M = cv2.getRotationMatrix2D((w / 2, h / 2), angle, 1)
rot_img = cv2.warpAffine(cv_img, M, (w, h))

# 82.1 ms ± 2.37 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
rot_img_pillow = np.array(pillow_img.rotate(angle))