Входной файл: | input.jpg | Ограничение времени: | 2 сек | |
Выходной файл: | output.txt | Ограничение памяти: | 512 Мб |
Требуется реализовать на языке Python набор функций, выполняющие этапы применения оператора Кэнни:
import numpy as np
from typing import Tuple
def gaussian_blur(img: np.ndarray, kernel: Tuple[int, int], sigma: float) -> np.ndarray:
'''Blurs an image using Gaussian filter.
Arguments:
img: input image, a 2d np.ndarray.
kernel: gaussian kernel size.
sigma: gaussian kernel standard deviation.
Returns:
Blurred image, a 2d np.ndarray of the same size and dtype as `img`.
'''
pass
def magnitude_and_direction(img: np.ndarray, kernel: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
'''Applies a filter to the image and computes magnitude and direction of the gradient.
The filter is applied using 'reflect' mode for border pixels, i.e. dcb|abcd|cba
Arguments:
img: input image, a 2d np.ndarray.
kernel: the filter kernel, a 2d np.ndarray with odd dimension sizes.
The kernel is applied over x dimension, kernel.T is applied over y
Returns:
Magnitude and direction of the gradient, two 2d np.ndarray objects of the same size and dtype as `img`.
The direction values lie in range [0, 2 * pi].
'''
pass
def edge_thinning(magnitude: np.ndarray, direction: np.ndarray) -> np.ndarray:
'''Performs edge thinning step of Canny algorithm using 0°, 45°, 90°, 135°, 180° (=0°) gradient direction
as described here https://en.wikipedia.org/wiki/Canny_edge_detector#Gradient_magnitude_thresholding_or_lower_bound_cut-off_suppression.
If the angle is equally close to two groups, the group with lower angle value is selected.
Arguments:
magnitude: magnitude of image gradient, a 2d np.ndarray.
direction: direction of image gradient, a 2d np.ndarray.
Returns:
Boolean mask of suppressed pixels (False if a pixel is suppresed, True if preserved), a 2d np.ndarray of the same size as `magnitude` and dtype bool.
'''
pass
def edge_tracking(magnitude: np.ndarray, mask: np.ndarray, low_threshold: float, high_threshold: float) -> np.ndarray:
'''Performs edge tracking step of Canny algorithm. The thresholds are inclusive.
Arguments:
magnitude: magnitude of image gradient, a 2d np.ndarray.
mask: pixel suppression mask, obtained by edge_thinning function.
low_threshold: weak pixel threshold.
high_threshold: strong pixel threshold.
Returns:
A 2d np.ndarray of the same size as `magnitude` and dtype bool, representing detected edges.
'''
pass
Код решения должен содержать только импортируемые модули, определение и реализацию функций.
№ | Входной файл (input.jpg ) |
Выходной файл (output.txt ) |
---|---|---|
1 |
input.jpg img = cv.imread('input.jpg')
img = cv.cvtColor(img, cv.COLOR_BGR2GRAY).astype(float)
blur = gaussian_blur(img, (5, 5), 1)
magnitude, direction = magnitude_and_direction(blur, np.array([[1, 0, -1], [2, 0, -2], [1, 0, -1]]))
mask = edge_thinning(magnitude, direction)
edges = edge_tracking(magnitude, mask, 0.1 * np.max(magnitude), 0.2 * np.max(magnitude))
cv.imwrite('sample.png', edges.astype(int) * 255)
|
sample.png |