imageSegmentation | Image Segmentation using Texture and Color | Computer Vision library

Since your question is not completely clear if you wanted to extract individual characters or whole words, here is the approach to do both.

Individual characters

The main idea here is

• Convert image to grayscale and gaussian blur
• Perform canny edge detection
• Find contours
• Iterate through contours and filter using a minimum area
• Obtain bounding boxes and extract ROI

Canny edge detection using cv2.Canny()

Now we iterate through contours using cv2.findContours() and filter using cv2.contourArea() then draw bounding boxes

Here's the results for some of your other input images