from cv2 import cv2

from flask import Flask, request, jsonify

import pytesseract

import re

from pytesseract import Output

from base64 import b64encode, b64decode

from json import dumps

import nltk

from nltk.tokenize import word_tokenize

import difflib

sm = difflib.SequenceMatcher(None)

# get grayscale image

def get_grayscale(image):

return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

#thresholding

def thresholding(image):

return cv2.threshold(image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]

def process_image(path=None,keyword=None,score=None):

score_data = 0

if score == None:

score_data = 0.1

else:

score_data = score

if path == None:

result = {'err': True, 'message': 'Path Can`t be empty', 'data':False}

code = 403

return result,code

if keyword == None:

result = {'err': True, 'message': 'keyword field required', 'data':False}

code = 403

return result,code

else:

img = cv2.imread(path)

overlay = img.copy()

gray = get_grayscale(img)

threshs = thresholding(gray)

custom_config = r'--psm 6'

resultOcr = pytesseract.image_to_string(threshs, output_type=Output.DICT)

d = pytesseract.image_to_data(img, output_type=Output.DICT)

text_ocr_result = resultOcr['text']

formatted_ocr_result = re.sub('[^a-zA-Z]', ' ', text_ocr_result)

formatted_ocr_result = re.sub(r'\s+', ' ', formatted_ocr_result)

sm.set_seq2(keyword.lower())

lasResultSimiliarity = []

for x in word_tokenize(formatted_ocr_result):

sm.set_seq1(x.lower())

if sm.ratio() >= float(score_data):

lasResultSimiliarity.append({x.lower():sm.ratio()})

result = {'text': resultOcr['text'], 'match_result': lasResultSimiliarity, 'image_result': ''}

all_word_length = len(d['text'])

for i in range(all_word_length):

match = re.search(r"\b%s\b" % keyword,d['text'][i].lower(),re.IGNORECASE)

if match:

(x, y, w, h) = (d['left'][i], d['top'][i], d['width'][i], d['height'][i])

img = cv2.rectangle(img, (x, y), (x + w, y + h), (64, 230, 61, 0.35), cv2.FILLED)

alpha = 0.4

image_new = cv2.addWeighted(overlay, alpha, img, 1 - alpha, 0)

_, buffer = cv2.imencode('.png', image_new)

png_as_text = b64encode(buffer)

base64_bytes = b64encode(png_as_text)

base64_string_last_result = b64decode(base64_bytes)

last_result = base64_string_last_result.decode('utf-8')

result['image_result'] = last_result

if not result['image_result']:

result = {'err': True, 'message':'keyword '+keyword+' no match!', 'data': False}

code = 404

return result,code

else:

result = {'err': False, 'message':'successfully recognized image', 'data': result}

code = 200

return result,code