Hi!
You can try this approach!
- Apply Binary thresholding to the image.
- Find contours with CV_RETR_LIST
- Now filter the contours based on contour Area.
- You can find the required two smaller contours marked as red.
- Now create new contour set named as vector<point> merged_contour_points;
Add these smaller contours in merged_contour_points.
e.g merged_contour_points.push_back(contours[i]);// filtered smaller contours.
- Find Convex hull for the Merged contour points.
Hope this helps for your problem.