如何在opencv中使用阈值获取白色字母

Question

I'm following this tutorial: https://www.smartspate.com/how-to-write-a-bot-in-python-for-online-games/

but I'm at the first part. He seems to threshold the image and get the white letters in the picture, but my thresholding does jack and I can't seem to get simple white letters without getting alot of the backround.

img = cv2.imread('bot_color.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
x, img = cv2.threshold(img, 150,255, cv2.THRESH_BINARY)
cv2.imshow('img', img)
cv2.waitKey(0)

My result with 252 threshold (exactly as his):

My result with a 127 threshold:

The pic I snipped from the site (original)

I snipped his pics using snippet. But even with other pics, I don't understand why I can't get simple white letters to appear (say for example, another game)

Answer 1

Like mentioned before, you likely do not have the "original photo" if it was not available for download from the site. This, however should not alter the way you detect the letters any more than simply adjusting your threshold values. Try adjusting the values to a very broad range before narrowing the range to cherrypick letters. Try something like this first:

    x, img = cv2.threshold(img, 75,255, cv2.THRESH_BINARY)

Then raise the 75 value or lower the 255 value to narrow the threshold.

Answer 2

As I looked at the site it seems that the original image is copyrighted and you cannot download the original. If you made a screenshot of the picture online a lot of information is probably lost and hence the difference in result. Open the image in paint and zoom in on the text and you will see the problem. Try doing the same with an image you make on your own. Cheers!

The mega pixel count is the total number of pixels in the image and is related to the number of pixels in the width and the height of the image. By resizing the size of the image reduce the mega pixel count of the photo so the PC can't store all the detail any more (there simply aren't enough pixels to do so). Thus removing details from the image.

EDIT:

I searched online for a screenshot of the same game and the threshold works okay.