skimage convert to grayscale resulting to a black image

Question

I am scraping image and trying to convert them into grayscale. But if i view the image. It is plain black. Can someone tell me what is wrong with this? Thanks

Here is my snippet

import cv2
from skimage import io
from skimage import data
from skimage.color import rgb2gray

#...some codes

for i in elements:
    try:
        i.click()
    except ElementClickInterceptedException:
        driver.execute_script("window.scrollTo(0, document.body.scrollHeight);")
        i.click()
    url = driver.find_element_by_xpath('//*[@class="slider-list"]/div/img').get_attribute('src').replace('height=600','height=1200').replace('quality=70','quality=100')
    print (url)
    name = driver.find_element_by_xpath('//*[@class="slider-list"]/div/img').get_attribute('alt') + '.jpg'

    print (name)
    img = io.imread(url)
    new = rgb2gray(img)
    cv2.imwrite(os.path.join(fldrnm, name), new)

Please remove the redundant logic for this issue and test it with minimal code. Simply File Read as Gray and then write. Also, you are missing the cv2 inclusion in the code presented. — Raviteja Narra, Feb 07 '20 at 03:05

Raviteja Narra · Accepted Answer · 2020-02-07T07:15:00.890

2

Import cv2
Img = cv2.inread(filepath)
Img = cv2.cvtColor(Img, cv2.COLOR_RGB2GRAY)
cv2.imwrite(path,img)

edited Feb 07 '20 at 07:15

answered Feb 07 '20 at 03:16

Raviteja Narra

458
3
11

I tried to change it to `new = rgb2gray(img)` and i got a black image result – draw134 Feb 07 '20 at 03:17
1

In the error you posted, the size seems to be weird. [164,165,151].... could you print the shape of img post reading. My guess this is not a 3 channel image. – Raviteja Narra Feb 07 '20 at 03:19
yah i have edited my question and i get rid of the error. Now i can save the image but the problem is when i view it, it is plain black – draw134 Feb 07 '20 at 03:20
Please do the following to debug. Print size of img, size of new, histogram of new. – Raviteja Narra Feb 07 '20 at 03:22
i tried `print(img.shape)` and this is the result `(735, 1200, 3)` what do you mean by this? – draw134 Feb 07 '20 at 03:31
And the size of new, its histogram? – Raviteja Narra Feb 07 '20 at 03:34
1

Maybe before saving the file, you can use cb2.imshow(...) to view it, and confirm the issue is not with rgb2grey() – Raviteja Narra Feb 07 '20 at 03:35
when i try to do `cv2.imshow('ImageWindow',new)` the window is not responding why is it happening? – draw134 Feb 07 '20 at 03:40
You shared the size of img. Please also share the size of new. – Raviteja Narra Feb 07 '20 at 03:42
the image showd is grayscale if i pud `cv2.waitKey(0` – draw134 Feb 07 '20 at 04:12
how to download this? if i dont put `cv2.waitKey(0` the downloaded images becomes pure black – draw134 Feb 07 '20 at 04:12
1

Cv2.waitKey is to be pasted after cv2.imshow call. Download what? So, with cv2.imshow you are able to see the gray scale image correctly? – Raviteja Narra Feb 07 '20 at 04:16
yes. but i got `cv2.imwrite(os.path.join(fldrnm, name), new)` to save the image.but the image saved is pure black. why is this happening? – draw134 Feb 07 '20 at 04:21
1

I am all out of options as well. The last thing would be to try to save the rgb image and check it. It might be the case that the cv2 module is not correctly installed. – Raviteja Narra Feb 07 '20 at 04:45
I can save the rgb image but after i add `new = rgb2gray(img)` and download it. it becomes pure black – draw134 Feb 07 '20 at 05:11
img = cv2.imread(r"FilePath") img = cv2.cvtColor(img,cv2.COLOR_RGB2GRAY) cv2.imwrite("temp.bmp",img) Last chance please try this. this is the code i usually use. – Raviteja Narra Feb 07 '20 at 06:56
i tried `img = cv2.imread(url)` returns an error, I only use `img = io.imread(url)` – draw134 Feb 07 '20 at 07:06

score 0 · Answer 2 · answered Feb 10 '20 at 11:55

The original cause of the problem is that skimage.color.rgb2gray will rescale an image to floating point values in [0, 1], even if the original image contains uint8 values in [0, 255]. This is done to preserve precision during long sequences of operations. Using skimage.io.imsave instead of cv2.imwrite in the original question would work, as well as using skimage.util.img_as_ubyte before saving.

See this page for more details:

https://scikit-image.org/docs/dev/user_guide/data_types.html

skimage convert to grayscale resulting to a black image

2 Answers2