segment non-homogeneous image background with opencv - python-3.x

Question

I have an image of a scanner and i wanted to segment the background but the background is not homogeneous. how could i do that?
My Input is a cloud points
I used the code below to convert to a depth image
import numpy as np
import cv2
import open3d as o3d
import sys
def convertToDepth():
argv = sys.argv[1]
argv2 = sys.argv[2]
pcd = o3d.io.read_point_cloud(argv)
p = np.asarray(pcd.points)
points = np.copy(p)
points = np.asarray(points, np.int16)
max_x = np.max(points[:,0])
min_x = np.min(points[:,0])
max_y = np.max(points[:,1])
min_y = np.min(points[:,1])
img = np.zeros(((max_x - min_x)+1,(max_y-min_y)+1), np.uint8)
img[img == 0] = 255
max_z = np.max(points[:,2])
min_z = np.min(points[:,2])
for p in points:
img[(p[0] - min_x),(p[1]-min_y)] = min(img[(p[0] - min_x),(p[1]-min_y)], int((p[2] - min_z)* (256/((max_z - min_z)+1))))
cv2.imwrite(argv2,img)
convertToDepth()
this is my depth image
I have tried this code to obtain the seeds
import cv2
import numpy as np
img = cv2.imread('amostra.png')
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(gray ,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
cv2.imwrite('thresh.png',thresh)
and obtained this
How can i get the seeds with an non-homogeneous background?
UPDATE:
I tried this code:
import cv2
import numpy as np
import sys
path_img = sys.argv[1]
img = cv2.imread(path_img)
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
for i in range(0,255):
thresh = gray.copy()
thresh[:] = 0
thresh[(gray == i)] = 255
num_labels, labels_im = cv2.connectedComponents(thresh)
areaMaior = 100
maskMaior = np.zeros((img.shape[0],img.shape[1]),np.uint8)
for j in range(1,num_labels):
mask = np.zeros((img.shape[0],img.shape[1]),np.uint8)
mask[labels_im==j] = 255
pixels = cv2.countNonZero(mask)
if(pixels > areaMaior):
maskMaior = mask
areaMaior = pixels
#cv2.imwrite('segment/areaimg'+str(i)+'.png', maskMaior)
gray[(maskMaior==255)] = 255
cv2.imwrite('gray12.png',gray)
And obtained this