FBL_threaded.py

from utils import *
import numpy as np

smoothed_image = 0

def FBL(image, etf, iterations = 1):
	
	global smoothed_image

	size = image.shape

	smoothed_image = image

	for i in range(iterations):

		H3 = np.zeros(size)

		for h in range(size[0]):
			
			print_progress('FBL S',h,size[0])

			for w in range(size[1]):
				total_weight = 0
				ang = angle(etf[h][w][0],etf[h][w][1])
				pix = flow_neighbour(ang*180/np.pi)
				for j in range(-5,6):
					if h + pix[0]*j < 0 or h + pix[0]*j >= size[0] or w + pix[1]*j < 0 or w + pix[1]*j >= size[1]: continue
					H3[h][w] += smoothed_image[h+pix[0]*j][w+pix[1]*j]*gaussian(abs(j),sig_spatial)*intensity_weight(smoothed_image[h][w],smoothed_image[h+pix[0]*j][w+pix[1]*j],sig_intensity1)
					total_weight += gaussian(abs(j),sig_spatial)*intensity_weight(smoothed_image[h][w],smoothed_image[h+pix[0]*j][w+pix[1]*j],sig_intensity1)
					
				H3[h][w] /= total_weight
		
		print ("")
		smoothed_image = np.zeros(size)

		for h in range(size[0]):
			
			print_progress('FBL T',h,size[0])

			for w in range(size[1]):
				total_weight = 0
				angle_perpendicular = (angle(etf[h][w][0],etf[h][w][1]) + np.pi/2)
				
				if angle_perpendicular > 2*np.pi:
					angle_perpendicular -= 2*np.pi
				pix = flow_neighbour(angle_perpendicular*180/np.pi)
				for j in range(-5,6):
					if h + pix[0]*j < 0 or h + pix[0]*j >= size[0] or w + pix[1]*j < 0 or w + pix[1]*j >= size[1]: continue
					smoothed_image[h][w] += H3[h+pix[0]*j][w+pix[1]*j]*gaussian(abs(j),sig_spatial)*intensity_weight(H3[h][w],H3[h+pix[0]*j][w+pix[1]*j],sig_intensity2)
					total_weight += gaussian(abs(j),sig_spatial)*intensity_weight(H3[h][w],H3[h+pix[0]*j][w+pix[1]*j],sig_intensity2)
					
				smoothed_image[h][w] /= total_weight

		print("")
	return smoothed_image