neighbour.py

import cv2
import numpy as np

import common

import math

def _get_neighbourhood(base, coords, radius=1):
    """Returns the neighbour of a given point in a specific radious"""
    x, y = coords
    ux, lx = x-radius if x-radius > 0 else 0, x+radius+1 if x+radius+1 < base.shape[0]-1 else base.shape[0]-1
    uy, ly = y-radius if y-radius > 0 else 0, y+radius+1 if y+radius+1 < base.shape[1]-1 else base.shape[1]-1
#    print("[{}:{}, {}:{}]".format(ux, lx, uy ,ly))
    return base[ux:lx, uy:ly]

def clean(input):
    """Performs a preliminary cleaing of the image using a custom filter based on neighbours"""
    result = np.zeros(input.shape, dtype=np.uint8)

    for c, v in np.ndenumerate(input):
         n = cv2.countNonZero(_get_neighbourhood(input, c, 3).ravel())
         if n > 9:
#             print("{} FG({})".format(c, n))
             result[c] = 255
    return result

def clean2(input):
    """Clean the inout image by performing a series of dilations and erosion operations"""
    dil_factor = int(math.floor(math.log10(abs(input.shape[0])))) + 2
    el = cv2.getStructuringElement(cv2.MORPH_RECT, (int(2*dil_factor + 1), int(2*dil_factor + 1)))
    kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (7, 6))
    
    input = cv2.medianBlur(input, 5)
    
    #input = cv2.dilate(input, el)
    input = cv2.morphologyEx(input, cv2.MORPH_CLOSE, kernel)
    
     
    return clean(input)
    


def imshow(img):
    plt.imshow(img)
    plt.show()

if __name__ == "__main__":
    i = cv2.imread("ref/gen/con/3.jpg")
    i = cv2.cvtColor(i, cv2.COLOR_BGR2GRAY)
    _, i = cv2.threshold(i, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)

    k1 = clean(i)
    k2 = clean2(i)
    cv2.imshow("Old", i)
    cv2.imshow("clean", k1)
    cv2.imshow("clean2", k2)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

    #_get_neighbourhood(k, (280,275))