处理图像以获得良好的OCR识别

Question

I'm trying to proccess a picture. Initially, there is a lot of noise, but I'm trying to make some improvements. Unfortunately, this does not give much result. It may be possible to realize such a thing that somehow select the center of each black line of letters in the word and do something like their skeleton. I have no idea how to do this, so please help here. The code i'm using now:

word = cv2.resize(word, (word.shape[1]*2, word.shape[0]*2))
kernel = np.array([[-1,-1,-1], [-1,9,-1], [-1,-1,-1]])
word = cv2.filter2D(word, -1, kernel)
word[np.where((word >= [180,180,180]).all(axis=2))] = [255,255,255]
word[np.where((word <= [179,179,179]).all(axis=2))] = [0,0,0]
cv2.imshow(str(i), word)

it gives this result: from this:

Answer 1

Try cv2.erode() .

cv2.bitwise_not(img,img)
kernel = np.ones((3,3),np.uint8)
erosion = cv2.erode(img,kernel,iterations = 1)