python - Image processing issues with blood vessels -

i'm trying extract blood vessels image, , so, i'm first equalizing image, applying clahe histogram obtain following result:

        clahe = cv2.createclahe(cliplimit=100.0, tilegridsize=(100,100))         self.cl1 = clahe.apply(self.result_array)         self.cl1 = 255 - self.cl1 

and i'm using otsu threshold extract blood vessels, failing well:

self.ret, self.thresh = cv2.threshold(self.cl1, 0,255,cv2.thresh_binary + cv2.thresh_otsu)         kernel = np.ones((1,1),np.float32)/1         self.thresh = cv2.erode(self.thresh, kernel, iterations=3)         self.thresh = cv2.dilate(self.thresh, kernel, iterations=3) 

here's result:

obviously there's lot of noise. i've tried using median blur, clusters noise , makes blob, in places. how go removing noise blood vessels?

this original image i'm trying extract blood vessels:

getting results difficult problem (you'll have somehow model structure of blood vessels , noise) can still better filtering.

one technique addressing kind of problems, inspired canny edge detector, using 2 thresholds - [hi,low] , classifying pixel p response r belonging blood vessel v if r > hi || (r > lo && 1 of p's neighbors in v).

also, when comes filtering, both bilateral filtering , meanshift filtering noisy images.

kernel3 = cv2.getstructuringelement(cv2.morph_ellipse,(3,3)) kernel5 = cv2.getstructuringelement(cv2.morph_ellipse,(5,5)) kernel7 = cv2.getstructuringelement(cv2.morph_ellipse,(7,7)) t_lo = 136 t_hi = 224  blured = cv2.pyrmeanshiftfiltering(img, 3, 9) #blured = cv2.bilateralfilter(img, 9, 32, 72)  clahe = cv2.createclahe(cliplimit=128.0, tilegridsize=(64, 64)) cl1 = clahe.apply(blured) cl1 = 255 - cl1  ret, thresh_hi = cv2.threshold(cl1, t_hi, 255, cv2.thresh_tozero) ret, thresh_lo = cv2.threshold(cl1, t_lo, 255, cv2.thresh_tozero) 

low threshold image hi threshold image

preparations , cleanup:

current = np.copy(thresh_hi) prev = np.copy(current) prev[:] = 0 current = cv2.morphologyex(current, cv2.morph_open, kernel5) iter_num = 0 max_iter = 1000 

not efficient way that... easy implement:

while np.sum(current - prev) > 0 , iter_num < max_iter:     iter_num = iter_num+1     prev = np.copy(current)     current = cv2.dilate(current, kernel3)     current[np.where(thresh_lo == 0)] = 0 

initial mask

remove small blobs:

contours, hierarchy = cv2.findcontours(current, cv2.retr_list, cv2.chain_approx_simple) contour in contours:     area = cv2.contourarea(contour)     if area < 256:         cv2.drawcontours( current, [contour], 0, [0,0,0], -1 ) 

after removing small blobs

morphological cleanup:

opening = cv2.morphologyex(current, cv2.morph_open, kernel7)    cl1[np.where(opening == 0)] = 0 

result

this no means optimal, think should provide enough tools start.