为什么使用cv2.imread从pdf文件读取Wand创建的图像为所有RGB返回全部255？

Question

I am trying to identify blobs of text in a pdf file. So say for example, there are different sections in an academic paper, and I want to identify the title as a section, the authors and addresses as a section, and the abstract as a section.

One solution i am thinking is to use cv2. I first convert pdf to a image using Wand using the following codes:

from wand.color import Color
from wand.image import Image as Img
with Img(filename='./files/paper.pdf', resolution=300) as img:
    img.background_color = Color("white")
    img.alpha_channel = 'remove'
    img.save(filename='test_file.jpg')

However, when I am trying to open the jpg file in cv2 with:

image = cv2.imread('test_file.jpg')

print image

the printout shows that all the values in that image is 255 for all pixels.

array([[[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       ...,

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]],

       [[255, 255, 255],
        [255, 255, 255],
        [255, 255, 255],
        ...,
        [255, 255, 255],
        [255, 255, 255],
        [255, 255, 255]]], dtype=uint8)

And then, when i want to use cv2.dnn.blobFromImage(), it just won't get it right.

What's going on? Was it because the pdf didn't get converted correctly into an image? But i tried

from PIL import Image
text = pytesseract.image_to_string(Image.open('test_file.jpg'))

, it returned all text to me...

Answer 1

See all the dots? The printing of the image is just showing a few pixels of the image. Assuming you have a pdf text document with a white background, it is safe to assume that all the edge pixels are white. The print will typically show you the corners of the image.

To show the image use

image = cv2.imread('test_file.jpg')
cv2.imshow('Image', image)
cv2.waitKey(0)

This will display the image in a window, and wait for you to press a key before disappearing.

Answer 2

Wand images are not numpy arrays and so cannot be simply opened in cv2. In Wand 5.3, there will be a way to import and export Wand images to and from numpy arrays.

In Wand 5.2, you can use import_pixels to convert a numpy array to a Wand image. In Wand 5.2, you can export a Wand image to numpy array that you should be able to use in cv2.

import numpy as np
from wand.image import Image

with Image(filename='rose.png') as img: 
    matrix = np.array(img)

matrix will be a numpy array that you should then be able to use in OpenCV