Any good OpenCV or SKImage techniques to thin out gridlines?

Question

I have extracted a clean grid pattern:

This above is the grid before I "skeletonize" (or thin, or perform the medial axis transform).

An below is the image after an application of skimage.skeletonize|medial_axis|thin or method=lee for the skeletonize:

These seem to eliminate the grid entirely due to the "boldness" or "thickness" of the lines.

Is there a preferred method to thin out these lines?

Answer 1

I have modified @Miki's answer (actually my search revealed that it was originally posted by another SO user in 2013). See if this solution is something that you could modify, by maybe tweaking a few parameters, to work for your case.

oElem = cv2.getStructuringElement(cv2.MORPH_RECT,(10,1))
h = cv2.morphologyEx(img, cv2.MORPH_OPEN, oElem,  iterations = 5)

oElem = cv2.getStructuringElement(cv2.MORPH_RECT,(1,10))
v = cv2.morphologyEx(img, cv2.MORPH_OPEN, oElem,  iterations = 5)

size = np.size(img)
skelh = np.zeros(img.shape,np.uint8)
skelv = np.zeros(img.shape,np.uint8)

ret,img = cv2.threshold(img,127,255,0)
element = cv2.getStructuringElement(cv2.MORPH_CROSS,(3,3))
done = False
while( not done):
    eroded = cv2.erode(h,element)
    temp = cv2.dilate(eroded,element)
    temp = cv2.subtract(h,temp)
    skelh = cv2.bitwise_or(skelh,temp)
    h = eroded.copy()
 
    if cv2.countNonZero(h)==0:
        done = True
        
done = False
while( not done):
    eroded = cv2.erode(v,element)
    temp = cv2.dilate(eroded,element)
    temp = cv2.subtract(v,temp)
    skelv = cv2.bitwise_or(skelv,temp)
    v = eroded.copy()
 
    if cv2.countNonZero(v)==0:
        done = True
        
skel = cv2.bitwise_or(skelh,skelv)