[FIXED] Pytesseract not recognising numbers

Issue

I'm trying to read numbers from a picture with pytesseract. This is my code:

import matplotlib.pyplot as plt
import pytesseract

img = cv2.imread(r'bild4.jpg', 2)

[ret, bw_img] = cv2.threshold(img, 200, 255, cv2.THRESH_BINARY)
imgplot = plt.imshow(bw_img)
plt.show()

text = pytesseract.image_to_string(bw_img, config='digits')
print("Text: " + text)

I tried many ways to preprocess the image, the best I got is: "673 504 . 5 552" Only the last line is correct. Without the config='digits' I get: "673 ost

Fir 504 .

5 552

ii"

I tried with black and withe only, witch is really easy to read for me as a human, but it don't recognice numbers at all...

Solution

You could use inRange thresholding

1. Read the image and convert it to the HSV color-scale

bgr = cv2.imread("JSe5v.jpg")
hsv = cv2.cvtColor(bgr, cv2.COLOR_BGR2HSV)

2. Initialize mask and kernel

mask = cv2.inRange(hsv, np.array([0, 0, 244]), np.array([179, 35, 255]))
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 3))

3. Remove the background using mask and kernel

dilated = cv2.dilate(mask, kernel, iterations=1)
thresh = cv2.bitwise_and(dilated, mask)

The result will be:

If you read it with psm mode 6

675 OS!
312504
5 552

1st line second part is not correct "51" is recognized as "S!"

You could look for improving the tesseract accuracy

Code:

import cv2
import pytesseract
import numpy as np

bgr = cv2.imread("JSe5v.jpg")
hsv = cv2.cvtColor(bgr, cv2.COLOR_BGR2HSV)
mask = cv2.inRange(hsv, np.array([0, 0, 244]), np.array([179, 35, 255]))
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 3))
dilated = cv2.dilate(mask, kernel, iterations=1)
thresh = cv2.bitwise_and(dilated, mask)

text = pytesseract.image_to_string(thresh, config="--psm 6")
print(text)

Answered By - Ahx