11. 이미지 변형(흐림)

In [1]:

from Custom.mediahelper import show_image_with_pil # 개발자 정의 모듈

11. 이미지 변형(흐림)¶

가우시안 블러¶

커널 사이즈 변화에 따른 흐림

• 커널 크기가 크면 더 넓은 영역의 픽셀을 평균화해 부드러운 블러 효과.

In [2]:

import cv2
img = cv2.imread('../Media/images/img.png')

_3 = (3,3)
_5 = (5,5)
_7 = (7,7)

kernel_3 = cv2.GaussianBlur(img, _3, 0) # 0 -> 표준편차
kernel_5 = cv2.GaussianBlur(img, _5, 0)
kernel_7 = cv2.GaussianBlur(img, _7, 0)

cv2.imshow('img', img)
cv2.imshow('kernel_3', kernel_3)
cv2.imshow('kernel_5', kernel_5)
cv2.imshow('kernel_7', kernel_7)

cv2.waitKey(0)
cv2.destroyAllWindows()

In [3]:

show_image_with_pil(img, 'img')
show_image_with_pil(kernel_3, 'kernel_3')
show_image_with_pil(kernel_5, 'kernel_5')
show_image_with_pil(kernel_7, 'kernel_7')

'img'

'kernel_3'

'kernel_5'

'kernel_7'

표준 편차 변화에 따른 흐림

• 표준편차가 클수록 블러 효과가 더 강해지며, 가우시안 필터의 분산이 증가.

In [4]:

import cv2
img = cv2.imread('../Media/images/img.png')

sigma_1 = cv2.GaussianBlur(img, (0,0), 1) # sigmaX
sigma_2 = cv2.GaussianBlur(img, (0,0), 2)
sigma_3 = cv2.GaussianBlur(img, (0,0), 3)
sigma_4 = cv2.GaussianBlur(img, (0,0), 4)

cv2.imshow('img', img)
cv2.imshow('sigma_1', sigma_1)
cv2.imshow('sigma_2', sigma_2)
cv2.imshow('sigma_3', sigma_3)
cv2.imshow('sigma_4', sigma_4)

cv2.waitKey(0)
cv2.destroyAllWindows()

In [5]:

show_image_with_pil(cv2.resize(img, None, fx = 0.5, fy = 0.5, interpolation = cv2.INTER_AREA), 'img')
show_image_with_pil(cv2.resize(sigma_1, None, fx = 0.5, fy = 0.5, interpolation = cv2.INTER_AREA), 'sigma_1')
show_image_with_pil(cv2.resize(sigma_2, None, fx = 0.5, fy = 0.5, interpolation = cv2.INTER_AREA), 'sigma_2')
show_image_with_pil(cv2.resize(sigma_3, None, fx = 0.5, fy = 0.5, interpolation = cv2.INTER_AREA), 'sigma_3')
show_image_with_pil(cv2.resize(sigma_4, None, fx = 0.5, fy = 0.5, interpolation = cv2.INTER_AREA), 'sigma_4')

'img'

'sigma_1'

'sigma_2'

'sigma_3'

'sigma_4'