Spaces:
Running
Running
| import cv2 | |
| import cv2 as cv | |
| import numpy as np | |
| import streamlit as st | |
| from PIL import Image | |
| import pytesseract | |
| def enhance(img): | |
| # Apply GaussianBlur to reduce noise | |
| blurred = cv2.GaussianBlur(img, (5, 5), 0) | |
| # Apply adaptive thresholding to enhance text | |
| thresh = cv2.adaptiveThreshold( | |
| blurred, | |
| 255, | |
| cv2.ADAPTIVE_THRESH_GAUSSIAN_C, | |
| cv2.THRESH_BINARY, | |
| 11, | |
| 2 | |
| ) | |
| # Perform morphological operations to remove small noise and connect text components | |
| kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3)) | |
| morph = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel) | |
| return morph | |
| act = [(150,240), (610,260)] | |
| act2 = [(130,440), (590,460)] | |
| acts=[act,act2] | |
| def align_images(ref_gray, input_gray, enh= False): | |
| """ | |
| Aligns the input image to the reference image using homography. | |
| Parameters: | |
| reference_image (numpy.ndarray): The reference image. | |
| input_image (numpy.ndarray): The input image to be aligned. | |
| Returns: | |
| numpy.ndarray: The aligned version of the input image. | |
| """ | |
| # # Convert images to grayscale | |
| # ref_gray = cv2.cvtColor(reference_image, cv2.COLOR_BGR2GRAY) | |
| # input_gray = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY) | |
| if enh: | |
| ref_gray = enhance(ref_gray) | |
| input_gray = enhance(input_gray) | |
| st.image(ref_gray) | |
| st.image(input_gray) | |
| # Detect ORB keypoints and descriptors | |
| orb = cv2.ORB_create(nfeatures=3000) | |
| keypoints1, descriptors1 = orb.detectAndCompute(ref_gray, None) | |
| keypoints2, descriptors2 = orb.detectAndCompute(input_gray, None) | |
| # Match descriptors using BFMatcher | |
| bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True) | |
| matches = bf.match(descriptors1, descriptors2) | |
| matches = sorted(matches, key=lambda x: x.distance) | |
| # Extract location of good matches | |
| ref_points = np.float32([keypoints1[m.queryIdx].pt for m in matches]).reshape(-1, 1, 2) | |
| input_points = np.float32([keypoints2[m.trainIdx].pt for m in matches]).reshape(-1, 1, 2) | |
| # Compute homography matrix | |
| H, mask = cv2.findHomography(input_points, ref_points, cv2.RANSAC, 5.0) | |
| # Warp input image to align with reference image | |
| height, width = ref_gray.shape | |
| aligned_image = cv2.warpPerspective(input_gray, H, (width, height)) | |
| return aligned_image | |
| def ocr_with_crop(aligned_image): | |
| # Open the image | |
| # img = Image.open(image_path) | |
| # img = cv2.imread(image_path,0) | |
| # img = enhance(img) | |
| # st.image(img) | |
| # st.write(type(img)) | |
| # enh = enhance(np.array(img)) | |
| # st.image(enh) | |
| # Define the coordinates for cropping | |
| def ocr(act): | |
| crop_coordinates = act | |
| # Convert to rectangular bounds (x1, y1, x2, y2) | |
| x1, y1 = crop_coordinates[0] | |
| x2, y2 = crop_coordinates[1] | |
| # Crop the image using the defined coordinates | |
| # cropped_img = img.crop((x1, y1, x2, y2)) | |
| cropped_img = aligned_image[y1:y2,x1:x2] | |
| st.image(cropped_img) | |
| # Perform OCR on the cropped image | |
| text = pytesseract.image_to_string(cropped_img) | |
| # Print the extracted text | |
| st.write(text) | |
| for cor in acts: | |
| ocr(cor) | |
| if __name__== "__main__": | |
| ref = cv.imread("r.png",0) | |
| if inp:= st.file_uploader("upload your form in image format", type=['png']): | |
| image = Image.open(inp) | |
| gray_image_pil = image.convert('L') | |
| image_array = np.array(gray_image_pil) | |
| st.image(image_array) | |
| align_image = align_images(ref,image_array) | |
| ocr_with_crop(align_image) | |