#coding: utf-8
import cv2
import numpy as np
from ..utils.generic_utils import handleKeyError, handleTypeError
OPENCV_BINARYZATIONS = {
"fixed" : None,
"mean" : cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
"gaussian" : cv2.ADAPTIVE_THRESH_MEAN_C,
}
[docs]def binarizer_creator(method, maxval=255, thresh=127, thresholdType=cv2.THRESH_BINARY, blockSize=11, const=2):
"""Returns a function which performs advanced morphological transformations.
Args:
method (str/int) : The identifier of binarization.
thresh (int) : Threshold value used in ``"fixed"`` binarizer.
maxval (int) : The maximum value.
thresholdType (int) : Thresholding type.
blockSize (odd) : Size of a pixel neighborhood that is used to calculate a threshold value for the pixel: ``3`` , ``5`` , ``7`` , and so on.
const (int) : Constant subtracted from the mean or weighted mean (see the details below). Normally, it is positive but may be zero or negative as well.
Returns:
function : Binarizer.
Examples:
>>> import cv2
>>> from pycharmers.opencv import binarizer_creator, SAMPLE_LENA_IMG, vconcat_resize_min, cv2plot
>>> from pycharmers.opencv.binary import OPENCV_BINARYZATIONS
>>> gray = cv2.imread(SAMPLE_LENA_IMG, 0)
>>> images = [gray]
>>> images.extend([binarizer_creator(key)(gray) for key in OPENCV_BINARYZATIONS.keys()])
>>> ax = cv2plot(vconcat_resize_min(*images), cmap="binary", figkeywargs={"figsize": (10,10)})
"""
handleTypeError(types=[str, int], method=method)
handleKeyError(lst=list(OPENCV_BINARYZATIONS.keys()), method=method)
if method == "fixed":
binarizer = lambda src, thresh=thresh, type=thresholdType : cv2.threshold(src=src, thresh=thresh, maxval=maxval, type=type)[1]
else:
adaptiveMethod = OPENCV_BINARYZATIONS.get(method, method)
binarizer = lambda src, blockSize=blockSize, C=const : cv2.adaptiveThreshold(src=src, maxValue=maxval, adaptiveMethod=adaptiveMethod, thresholdType=thresholdType, blockSize=blockSize, C=C)
return binarizer
[docs]def findBiggestContour(contours, eta=0.1):
"""
Args:
contours (list) : Contours. (e.g. The return of ``cv2.findContours`` )
Returns:
biggest (np.ndarray) : Countour of the biggest area.
max_area (int) : Area of the biggest area.
Examples:
>>> import cv2
>>> from pycharmers.opencv import findBiggestContour, SAMPLE_LENA_IMG
>>> img_gray = cv2.imread(SAMPLE_LENA_IMG, 0)
>>> img_th = cv2.Canny(img_gray, 100, 100)
>>> contours, hierarchy = cv2.findContours(image=img_th, mode=cv2.RETR_EXTERNAL, method=cv2.CHAIN_APPROX_SIMPLE)
>>> biggest, max_area = findBiggestContour(contours, eta=0.2)
>>> max_area
0 # There is no closed countour.
"""
biggest = np.array([])
max_area = 0
for contour in contours:
area = cv2.contourArea(contour)
epsilon = eta * cv2.arcLength(curve=contour, closed=True)
approx = cv2.approxPolyDP(curve=contour, epsilon=epsilon, closed=True)
if area > max_area and len(approx) == 4:
biggest = approx
max_area = area
return biggest, max_area
[docs]def reorder_contour(points):
"""Reorder the points created by ``cv2.approxPolyDP``
Args:
points (np.ndarray) : Points.
Returns:
ret (np.ndarray) : Reordered Points.
"""
points = points.reshape((4, 2))
ret = np.zeros((4, 1, 2), dtype=np.int32)
sum = np.sum(points, axis=1)
diff = np.diff(points, axis=1)
ret[0] = points[np.argmin(sum)]
ret[1] = points[np.argmin(diff)]
ret[2] = points[np.argmax(diff)]
ret[3] = points[np.argmax(sum)]
return ret
