python，opencv，图像数组到二进制

Question

I have a large image , using cv2 module in python and some coordinates i cropped the image:

  img = cv.imread(image_path)
  crop_img = img[y1:y2,x1:x2]
  cv.imwrite(cropPath, crop_img)

now the crop_img is a numpy.ndarray type. then I save this image to disk and read its contents in a binary format using an open() function

  with open(cropPath, 'rb') as image_file:
    content = image_file.read()

and I get the binary representation. Is there any way to do the above operations without saving the image to disk. Not saving to disk will save a lot of time, I am not able to find any method to do this. if anyone could point in the right direction, that would be helpful.

Answer 1

found the answer on this thread: Python OpenCV convert image to byte string?

converting a image represented through a numpy array into string can be done by using imencode and tostring functions in cv2

>>> img_str = cv.imencode('.jpg', img)[1].tostring()
>>> type(img_str)
 'str'

Answer 2

If you use cv2.imwrite() ， then you will get an image in image format,such as png, jpg, bmp and so on. Now if you open(xxx,"rb") as a normal binary file, it will go wrong, because it is AN IMAGE in IMAGE FILE FORMAT .

The simplest way is use np.save() to save the np.ndarray to the disk ( serialize ) in .npy format. The use np.load() to load from disk ( deserialize ).

An alternative is pickle.dump()/pickle.load() .

Here is an example:

#!/usr/bin/python3
# 2017.10.04 21:39:35 CST

import pickle 
imgname = "Pictures/cat.jpg"

## use cv2.imread()/cv2.imwrite() 
img = cv2.imread(imgname)

## use np.save() / np.load()
np.save(open("another_cat1.npy","wb+"), img)
cat1 = np.load(open("another_cat1.npy","rb"))

## use pickle.dump() / pickle.load()
pickle.dump(img, open("another_cat2.npy","wb+"))
cat2 = pickle.load(open("another_cat2.npy", "rb"))

cv2.imshow("img", img);
cv2.imshow("cat1", cat1);
cv2.imshow("cat2", cat2);
cv2.waitKey();cv2.destroyAllWindows()

The result: