ways to reduce noise + smooth the image to create a mask for the largest object

Question

I need your advice, guys! So I am trying to create a mask for a football (i.e. soccer) pitch, and if I only use filtering by the most common hue/saturation/value values, I get the following image

As you can see, the desired part is inside the grey boundary I drawn and there is a lot of noise here - trees, lines, external objects, and of course I would like to get rid of it. The desired outcome is something similar to this:

I thought about an algorithm that would transform the first image into another by analyzing each pixel's surrounding and color it white if more than threshold% of pixels into a (x, x) square is white, otherwise black.

Do you know if there is an implementation on openCV or similar libraries for this or I should build it from scratch?

Also, maybe you can propose other way to deal with the noise and external objects? I already tried the morphological transform and blurring techniques, but either I don't do it right or it doesn't work well for my problem.

Thank you in advance for your advice!

Answer 1

I actually found an easy implementation of the algo I proposed - I simply use cv2.blur on the image and then filter with cv2.inRange, so it does exactly what I wanted it to do.