• 1. PIL (Pillow): Open an Image
• 2. OpenCV: Change an Image
• 3. Tessereact (PyTessereact): OCR an Image

Install: https://pillow.readthedocs.io/en/stable/installation/basic-installation.html

pip install --upgrade Pillow

Document: https://pillow.readthedocs.io/en/stable/reference/Image.html

Install: https://pypi.org/project/opencv-python

pip install opencv-python

Install Tessereact: https://guides.library.illinois.edu/c.php?g=347520&p=4121425

% brew install tesseract-lang
 
% tesseract --version
tesseract 5.4.1
 leptonica-1.84.1
  libgif 5.2.1 : libjpeg 8d (libjpeg-turbo 3.0.0) : libpng 1.6.43 : libtiff 4.6.0 : zlib 1.2.12 : libwebp 1.4.0 : libopenjp2 2.5.2
 Found NEON
 Found libarchive 3.7.4 zlib/1.2.12 liblzma/5.4.6 bz2lib/1.0.8 liblz4/1.9.4 libzstd/1.5.6
 Found libcurl/8.7.1 SecureTransport (LibreSSL/3.3.6) zlib/1.2.12 nghttp2/1.61.0
(jupyter) tungnt@MacBook-Pro-cua-Nguyen-2 jupyter % 

Install pytesseract: https://pypi.org/project/pytesseract

pip install pytesseract

Language:

• https://tesseract-ocr.github.io/tessdoc/Data-Files-in-different-versions.html
• https://www.kaggle.com/code/dhorvay/pytesseract-multiple-languages

print(pytesseract.get_languages(config='.'))
 
['afr', 'amh', 'ara', 'asm', 'aze', 'aze_cyrl', 'bel', 'ben', 'bod', 'bos', 'bre', 'bul', 'cat', 'ceb', 'ces', 'chi_sim', 'chi_sim_vert', 'chi_tra', 'chi_tra_vert', 'chr', 'cos', 'cym', 'dan', 'deu', 'div', 'dzo', 'ell', 'eng', 'enm', 'epo', 'equ', 'est', 'eus', 'fao', 'fas', 'fil', 'fin', 'fra', 'frk', 'frm', 'fry', 'gla', 'gle', 'glg', 'grc', 'guj', 'hat', 'heb', 'hin', 'hrv', 'hun', 'hye', 'iku', 'ind', 'isl', 'ita', 'ita_old', 'jav', 'jpn', 'jpn_vert', 'kan', 'kat', 'kat_old', 'kaz', 'khm', 'kir', 'kmr', 'kor', 'kor_vert', 'lao', 'lat', 'lav', 'lit', 'ltz', 'mal', 'mar', 'mkd', 'mlt', 'mon', 'mri', 'msa', 'mya', 'nep', 'nld', 'nor', 'oci', 'ori', 'osd', 'pan', 'pol', 'por', 'pus', 'que', 'ron', 'rus', 'san', 'sin', 'slk', 'slv', 'snd', 'snum', 'spa', 'spa_old', 'sqi', 'srp', 'srp_latn', 'sun', 'swa', 'swe', 'syr', 'tam', 'tat', 'tel', 'tgk', 'tha', 'tir', 'ton', 'tur', 'uig', 'ukr', 'urd', 'uzb', 'uzb_cyrl', 'vie', 'yid', 'yor']
 
from langdetect import detect_langs 
 
detect_langs(ocr_result_original)
 
tesseract --list-langs

https://github.com/1sitevn/python-jupyter/blob/main/ocr/01_OCR_Pillow.ipynb

import cv2
import pytesseract
from PIL import Image
 
image_path = "../data/ocr/page_01.jpg"
 
image = Image.open(image_path)
 
print(image.size)
 
image.rotate(90).show()
 
image.save("../data/temp/page_01.jpg")

https://github.com/1sitevn/python-jupyter/blob/main/ocr/02_OCR_Preprocess_Images.ipynb

import cv2
import pytesseract
import numpy as np
from PIL import Image
from matplotlib import pyplot as plt

image_file = "../data/ocr/page_01.jpg"
 
img = cv2.imread(image_file)
 
def display(im_path):
    dpi = 80
    im_data = plt.imread(im_path)
 
    height, width  = im_data.shape[:2]
 
    # What size does the figure need to be in inches to fit the image?
    figsize = width / float(dpi), height / float(dpi)
 
    # Create a figure of the right size with one axes that takes up the full figure
    fig = plt.figure(figsize=figsize)
    ax = fig.add_axes([0, 0, 1, 1])
 
    # Hide spines, ticks, etc.
    ax.axis('off')
 
    # Display the image.
    ax.imshow(im_data, cmap='gray')
 
    plt.show()
 
display(image_file)    

inverted_image = cv2.bitwise_not(img)
 
cv2.imwrite("../data/temp/inverted.jpg", inverted_image)
 
display("../data/temp/inverted.jpg")

def grayscale(image):
    return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
 
gray_image = grayscale(img)
 
cv2.imwrite("../data/temp/gray.jpg", gray_image)    
 
display("../data/temp/gray.jpg")

thresh, im_bw = cv2.threshold(gray_image, 210, 230, cv2.THRESH_BINARY)
 
cv2.imwrite("../data/temp/bw_image.jpg", im_bw)
 
display("../data/temp/bw_image.jpg")

def noise_removal(image):
    kernel = np.ones((1, 1), np.uint8)
    image = cv2.dilate(image, kernel, iterations=1)
    kernel = np.ones((1, 1), np.uint8)
    image = cv2.erode(image, kernel, iterations=1)
    image = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)
    image = cv2.medianBlur(image, 3)
    return (image)
 
no_noise = noise_removal(im_bw)
 
cv2.imwrite("../data/temp/no_noise.jpg", no_noise)
 
display("../data/temp/no_noise.jpg")    

def thin_font(image):
    image = cv2.bitwise_not(image)
    kernel = np.ones((2,2),np.uint8)
    image = cv2.erode(image, kernel, iterations=1)
    image = cv2.bitwise_not(image)
    return (image)
 
eroded_image = thin_font(no_noise)
 
cv2.imwrite("../data/temp/eroded_image.jpg", eroded_image)
 
display("../data/temp/eroded_image.jpg")      

def thick_font(image):
    image = cv2.bitwise_not(image)
    kernel = np.ones((2,2),np.uint8)
    image = cv2.dilate(image, kernel, iterations=1)
    image = cv2.bitwise_not(image)
    return (image)
 
dilated_image = thick_font(no_noise)
 
cv2.imwrite("../data/temp/dilated_image.jpg", dilated_image) 
 
display("../data/temp/dilated_image.jpg")    

new = cv2.imread("../data/ocr/page_01_rotated.JPG")
 
display("../data/ocr/page_01_rotated.JPG")
 
def getSkewAngle(cvImage) -> float:
    # Prep image, copy, convert to gray scale, blur, and threshold
    newImage = cvImage.copy()
    gray = cv2.cvtColor(newImage, cv2.COLOR_BGR2GRAY)
    blur = cv2.GaussianBlur(gray, (9, 9), 0)
    thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
 
    # Apply dilate to merge text into meaningful lines/paragraphs.
    # Use larger kernel on X axis to merge characters into single line, cancelling out any spaces.
    # But use smaller kernel on Y axis to separate between different blocks of text
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 5))
    dilate = cv2.dilate(thresh, kernel, iterations=2)
 
    # Find all contours
    contours, hierarchy = cv2.findContours(dilate, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
    contours = sorted(contours, key = cv2.contourArea, reverse = True)
    for c in contours:
        rect = cv2.boundingRect(c)
        x,y,w,h = rect
        cv2.rectangle(newImage,(x,y),(x+w,y+h),(0,255,0),2)
 
    # Find largest contour and surround in min area box
    largestContour = contours[0]
    print (len(contours))
    minAreaRect = cv2.minAreaRect(largestContour)
    cv2.imwrite("temp/boxes.jpg", newImage)
    # Determine the angle. Convert it to the value that was originally used to obtain skewed image
    angle = minAreaRect[-1]
    if angle < -45:
        angle = 90 + angle
    return -1.0 * angle
 
# Rotate the image around its center
def rotateImage(cvImage, angle: float):
    newImage = cvImage.copy()
    (h, w) = newImage.shape[:2]
    center = (w // 2, h // 2)
    M = cv2.getRotationMatrix2D(center, angle, 1.0)
    newImage = cv2.warpAffine(newImage, M, (w, h), flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_REPLICATE)
    return newImage
 
# Deskew image
def deskew(cvImage):
    angle = getSkewAngle(cvImage)
    return rotateImage(cvImage, -1.0 * angle)
 
fixed = deskew(new)
 
cv2.imwrite("../data/temp/rotated_fixed.jpg", fixed)
 
display("../data/temp/rotated_fixed.jpg")       

display("../data/temp/no_noise.jpg")
 
def remove_borders(image):
    contours, heiarchy = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    cntsSorted = sorted(contours, key=lambda x:cv2.contourArea(x))
    cnt = cntsSorted[-1]
    x, y, w, h = cv2.boundingRect(cnt)
    crop = image[y:y+h, x:x+w]
    return (crop)
 
no_borders = remove_borders(no_noise)
 
cv2.imwrite("../data/temp/no_borders.jpg", no_borders)
 
display('../data/temp/no_borders.jpg')       

color = [255, 255, 255]
top, bottom, left, right = [150]*4
 
image_with_border = cv2.copyMakeBorder(no_borders, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)
 
cv2.imwrite("../data/temp/image_with_border.jpg", image_with_border)
 
display("../data/temp/image_with_border.jpg")   

https://github.com/1sitevn/python-jupyter/blob/main/ocr/03_OCR_Introduction_To_PyTesseract.ipynb

import cv2
import pytesseract
from PIL import Image
 
img_file = "../data/ocr/page_01.jpg"
no_noise = "../data/temp/no_noise.jpg"
 
img1 = Image.open(img_file)
 
display(img1)
 
ocr_result1 = pytesseract.image_to_string(img1)
 
print(ocr_result1)
 
img2 = Image.open(no_noise)
 
display(img2)
 
ocr_result2 = pytesseract.image_to_string(img2)
 
print(ocr_result2)
 
"""
“GABRIEL Meamall
 
On Easter movning in the year 1944, I took my six-year-old
son by the hand and began walking fron my home town toward the
valleys and forests of the Carpathizn mountains. For nearly
eight months we lived in barns, attics and makeshift eabins. With
the gene nous help of an unusually courageous man, we managed to
survive Europe's greatest fit of madness. Those who walked in
the opposite direction on that Easter day were lese fortunate.
They were taken in trainloads to places whose once obscure names
are now, and forever will be, synonymous with terror, evil and
death. What follows is our story of survival told to the best
of my ability, in plain, simple language.
 
In March of 1944 the SS troops took over the internal affairs
of Hungary and proceeded to organize the deportation of the dows.
To the Nazie thie was a routine assignment; within hours all local
officials were informed of operational plans. The high command
issued a directive designed to placate Jewish fears and induce
cooperation. It was announced that the Jews would be shipped to
Poland as an emergency labor force and that they were only being
drafted for temporary work. There were many who believed this
version. Others, less credulous, resigred themselves and hoped for
the vest. Still others began to make plans for escape. By Aprii
13 the Hungarian Jews were being rounded up from all over the .
country in what was once a huge brick factory. The rest is well
known. :
 
J was working in Ungvar and usually came home on weekends.
 
At that time it was no longer possible for a Jew to travel freely.
"""

https://github.com/1sitevn/python-jupyter/blob/main/ocr/04_OCR_an_Index_with_PyTesseract.ipynb

import cv2
import pytesseract
from PIL import Image
 
image_file = "../data/ocr/index_02.jpg"
 
img = Image.open(image_file)
 
ocr_result = pytesseract.image_to_string(img)
 
print(ocr_result)
 
lines = ocr_result.split("\n\n")
 
for line in lines:
    temp_line = line.replace(",", "")
    if temp_line.isdigit():
        pass
    else:
        components = []
        segs = line.split(",")
        for seg in segs:
            seg = seg.strip()
            num = False
            for character in seg:
                if character.isdigit():
                    num = True
            if num == False:
                components.append(seg)
        print (components)

https://github.com/1sitevn/python-jupyter/blob/main/ocr/05_OCR_Bounding_Boxes_with_OpenCV.ipynb

import cv2
import pytesseract
from PIL import Image
 
image = cv2.imread("../data/ocr/index_02.JPG")
 
base_image = image.copy()
 
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
cv2.imwrite("../data/temp/index_gray.png", gray)
 
blur = cv2.GaussianBlur(gray, (7,7), 0)
cv2.imwrite("../data/temp/index_blur.png", blur)
 
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
cv2.imwrite("../data/temp/index_thresh.png", thresh)
 
kernal = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 13))
cv2.imwrite("../data/temp/index_kernal.png", kernal)
 
dilate = cv2.dilate(thresh, kernal, iterations=1)
cv2.imwrite("../data/temp/index_dilate.png", dilate)
 
cnts = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cents[1]
cnts = sorted(cnts, key=lambda x: cv2.boundingRect(x)[0])
 
results  = []
for c in cnts:
    x, y, w, h = cv2.boundingRect(c)
    if h > 200 and w > 20:
        roi = image[y:y+h, x:x+h]
        cv2.rectangle(image, (x, y), (x+w, y+h), (36, 255, 12), 2)
        ocr_result = pytesseract.image_to_string(roi)
        ocr_result = ocr_result.split("\n")
        for item in ocr_result:
            results.append(item)
 
cv2.imwrite("../data/temp/index_bbox_new.png", image)
 
print (results)

https://github.com/1sitevn/python-jupyter/blob/main/ocr/06_OCR_a_Text_with_Marginalia_by_Extracting_the_Body.ipynb

import cv2
import pytesseract
from PIL import Image
 
image = cv2.imread("../data/ocr/sample_mgh.JPG")
 
base_image = image.copy()
 
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (7,7), 0)
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
 
kernal = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 50))
dilate = cv2.dilate(thresh, kernal, iterations=1)
 
cv2.imwrite("../data/temp/sample_dilated.png", dilate)
 
cnts = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
cnts = sorted(cnts, key=lambda x: cv2.boundingRect(x)[1])
 
for c in cnts:
    x,y,w,h = cv2.boundingRect(c)
    if h > 200 and w > 250:
        roi = base_image[y:y+h, x:x+w]
        cv2.rectangle(image, (x,y), (x+w, y+h), (36, 255, 12), 2)
 
cv2.imwrite("../data/temp/sample_boxes.png", image)
 
ocr_result_original = pytesseract.image_to_string(base_image)
 
print(ocr_result_original)
 
ocr_result_new = pytesseract.image_to_string(roi)
 
print(ocr_result_new)

https://github.com/1sitevn/python-jupyter/blob/main/ocr/07_OCR_Separate_a_Footnote_from_Body_Text.ipynb

import cv2
import pytesseract
from PIL import Image
 
image = cv2.imread('../data/ocr/sample_mgh_2.jpg')
base_image = image.copy()
 
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (7,7), 0)
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
 
# Create rectangular structuring element and dilate
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,25))
dilate = cv2.dilate(thresh, kernel, iterations=1)
 
# Find contours and draw rectangle
cnts = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
cnts = sorted(cnts, key=lambda x: cv2.boundingRect(x)[1])
main_text = ""
for c in cnts:
    x,y,w,h = cv2.boundingRect(c)
    if h > 200 and w > 250:
        roi = base_image[y:y+h, 0:x]
#         cv2.rectangle(image, (0, y), (x, 0 + h+20), (36,255,12), 2)
 
        constant= cv2.copyMakeBorder(roi.copy(),30,30,30,30,cv2.BORDER_CONSTANT,value=[255,255,255])
        ocr_result = pytesseract.image_to_string(constant)
        cv2.imwrite("../data/temp/output.png", roi)
 
        print (ocr_result)
#         print (ocr_result)
# cv2.imwrite("temp/output.png", image)

• OCR in Python Tutorials
• https://github.com/wjbmattingly/ocr_python_textbook
• https://www.affinda.com/tech-ai/how-to-convert-image-to-text-using-python
• https://tesseract-ocr.github.io/tessdoc/Compiling.html#macos
• https://nanonets.com/blog/ocr-with-tesseract/