Text extraction with OCR

Question

The provided image has various elements of a monitor. I want to extract the text of each of them. I have tried various approaches, but there is a lot of text that I cannot identify.enter image description here

1. using easyocr
2. using tesseract
3. or, the image has squares of different shapes. I have also tried to identify the rectangles and then extract the image from the identified rectangles.

• Sharpen the edges of a square. Load an image, convert it to grayscale, smooth the middle blur, and sharpen to enhance the edges.
• Obtain a binary image and remove noise. Set the threshold to get a black and white binary image. Depending on the image, either Otsu's thresholding or adaptive thresholding will work. Here we create a rectangular kernel and perform a morphological transformation to remove noise and enhance square contours
• Detect and extract the rectangle. Next, we find the contours and filter them using a min/max threshold region. All contours that pass through the filter will be squares, so we get the bounding rectangle coordinates to extract each ROI, create a green rectangular box in the image, and use the final image low enter image description here enter image description here

---

Below is the logic used

import cv2
import numpy as np
from PIL import Image

def integrated_preprocess_image_with_size_filter(input_file_path, output_file_path):
    
    img = Image.open(input_file_path)
    
    
    if img.mode == 'RGBA':
        img = img.convert('RGB')

    
    dpi = img.info['dpi'] if 'dpi' in img.info else (72, 72)

    
    if dpi[0] < 300 or dpi[1] < 300:
        img.save(input_file_path, dpi=(300, 300))
        img = Image.open(input_file_path)
    
    
    img_cv = np.array(img)
    img_cv = cv2.cvtColor(img_cv, cv2.COLOR_RGB2BGR)

    
    gray_image = cv2.cvtColor(img_cv, cv2.COLOR_BGR2GRAY)

    
    blur_kernel_size = 3 if np.var(gray_image) < 100 else 3

    
    blurred_image = cv2.bilateralFilter(gray_image, blur_kernel_size, 75, 75)

    
    sharpening_kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]])
    sharpened_image = cv2.filter2D(blurred_image, -1, sharpening_kernel)

    
    _, binary_image = cv2.threshold(sharpened_image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

    
    kernel_size = int(np.mean(gray_image) // 50000) + 1
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_size, kernel_size))
    opened_image = cv2.morphologyEx(binary_image, cv2.MORPH_OPEN, kernel)

    
    contours, _ = cv2.findContours(opened_image, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
    for cnt in contours:
        epsilon = 0.02 * cv2.arcLength(cnt, True)
        approx = cv2.approxPolyDP(cnt, epsilon, True)

        if len(approx) == 4 and 100 < cv2.contourArea(cnt) < 50000:
            x, y, w, h = cv2.boundingRect(approx)
            
            if w > 90 and h > 90:
                cv2.rectangle(img_cv, (x, y), (x+w, y+h), (0, 255, 0), 2)

    
    cv2.imwrite(output_file_path, img_cv)


integrated_preprocess_image_with_size_filter('./test_img.png', './path_to_output.jpg')

---

To approach it in different ways, my approach was to extract the rectangular boxes and then proceed with the OCR.

I want to know how to extract text from that image in the best way possible.

Thank you.