[OpenCV/Python] 임계값(Threshold) 처리 (BINARY, BINARY_INV, TRIANGLE, OTSU 등)

임계값 처리 (Threshold)

• cv2.THRESH_BINARY
• cv2.THRESH_BINARY_INV
• cv2.THRESH_TRUNC
• cv2.THRESH_TOZERO
• cv2.THRESH_TOZERO_INV

Parameters

src	input array (multiple-channel, 8-bit or 32-bit floating point).
dst	output array of the same size and type and the same number of channels as src.
thresh	threshold value.
maxval	maximum value to use with the THRESH_BINARY and THRESH_BINARY_INV thresholding types.
type	thresholding type.

# lena image
src = cv2.imread('./lena.jpg', cv2.IMREAD_GRAYSCALE)
ret, dst = cv2.threshold(src, 100, 255, cv2.THRESH_BINARY) #threshold=100, THRESH_BINARY
print('ret=', ret)

cv2.imshow('src', src)
cv2.imshow('dst', dst)
cv2.waitKey()
cv2.destroyAllWindows()
# ret= 100.0

# THRESH_BINARY vs THRESH_BINARY_INV

# cup image
src = cv2.imread('./cup.png', cv2.IMREAD_GRAYSCALE)
ret1, dst1 = cv2.threshold(src, 183, 255, cv2.THRESH_BINARY) #threshold=183
ret2, dst2 = cv2.threshold(src, 180, 255, cv2.THRESH_BINARY_INV) #threshold=180

cv2.imshow('src', src)
cv2.imshow('dst1', dst1)
cv2.imshow('dst2', dst2)
cv2.waitKey()
cv2.destroyAllWindows()

 

# THRESH_TRUNC vs THRESH_TOZERO vs THRESH_TOZERO_INV

src = cv2.imread('./cup.png', cv2.IMREAD_GRAYSCALE)
ret1, dst1 = cv2.threshold(src, 180, 255, cv2.THRESH_TRUNC)
ret2, dst2 = cv2.threshold(src, 180, 255, cv2.THRESH_TOZERO)
ret3, dst3 = cv2.threshold(src, 180, 255, cv2.THRESH_TOZERO_INV)

cv2.imshow('dst1', dst1)
cv2.imshow('dst2', dst2)
cv2.imshow('dst3', dst3)
cv2.waitKey()
cv2.destroyAllWindows()

 

# openCV 로고 이미지

src = cv2.imread('./opencv_logo.png', cv2.IMREAD_GRAYSCALE) 
ret, dst = cv2.threshold(src, 240, 255, cv2.THRESH_BINARY)

cv2.imshow('src', src)
cv2.imshow('dst', dst)
cv2.waitKey()
cv2.destroyAllWindows()
# 흰 배경에 있는 도형들이라 구분이 잘 됨

# 정규화 (normarlization)

• minMaxLoc() : 최대값, 최소값, 최대값 위치, 최소값 위치 반환

# minMaxLoc

src = cv2.imread('./lena.jpg', 0)

minVal, maxVal, minLoc, maxLoc = cv2.minMaxLoc(src)

print('src')
print('minVal:', minVal)
print('maxVal:', maxVal)
print('minLoc:', minLoc)
print('maxLoc:', maxLoc)
print('src[minLoc[1], minLoc[0]]:', src[minLoc[1], minLoc[0]])
print('src[maxLoc[1], maxLoc[0]]:', src[maxLoc[1], maxLoc[0]])

# src
# minVal: 18.0
# maxVal: 247.0
# minLoc: (265, 198)
# maxLoc: (116, 273)
# src[minLoc[1], minLoc[0]]: 18
# src[maxLoc[1], maxLoc[0]]: 247

 

• normalize

Parameters

src	input array.
dst	output array of the same size as src .
alpha	norm value to normalize to or the lower range boundary in case of the range normalization.
beta	upper range boundary in case of the range normalization; it is not used for the norm normalization.
norm_type	normalization type .
dtype	when negative, the output array has the same type as src; otherwise, it has the same number of channels as src and the depth =CV_MAT_DEPTH(dtype).
mask	optional operation mask.

# normalize

dst = cv2.normalize(src, None, 0, 255, cv2.NORM_MINMAX)
minVal, maxVal, minLoc, maxLoc = cv2.minMaxLoc(dst)

print('src')
print('minVal:', minVal)
print('maxVal:', maxVal)
print('minLoc:', minLoc)
print('maxLoc:', maxLoc)
print('src[minLoc[1], minLoc[0]]:', src[minLoc[1], minLoc[0]])
print('src[maxLoc[1], maxLoc[0]]:', src[maxLoc[1], maxLoc[0]])

# src
# minVal: 0.0
# maxVal: 255.0
# minLoc: (265, 198)
# maxLoc: (116, 273)
# src[minLoc[1], minLoc[0]]: 18
# src[maxLoc[1], maxLoc[0]]: 247

# 이미지 확인
cv2.imshow('src', src)
cv2.imshow('dst', dst)
cv2.waitKey()
cv2.destroyAllWindows()
# 조금 더 선명해짐 (밝기 대비가 강해짐)

 

• 0~255까지 정규화 직접 만들어서 이미지 구현

src = cv2.imread('./lena.jpg', 0)

print('min:', src.min()) # min: 18
print('max:',src.max()) # max: 247

dst = (src - np.min(src)) / (np.max(src)-np.min(src))*255  #정규화

print('min:',dst.min()) # min: 0.0
print('max:',dst.max()) # max: 255.0

# 결과 확인
cv2.imshow('src', src)
cv2.imshow('dst', np.uint8(dst))
cv2.waitKey()
cv2.destroyAllWindows()

 

 

# 자동 임계값 계산

• cv2.THRESH_TRIANGLE : 삼각 알고리즘을 이용한 자동 임계값 결정
• cv2.THRESH_OTSU : 오츠 알고리즘을 이용한 자동 임계값 결정

# lena image
src = cv2.imread('./lena.jpg', 0)

ret, dst = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY)
ret1, dst1 = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY+cv2.THRESH_TRIANGLE)
ret2, dst2 = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)
print('threshold=', ret)
print('threshold_triangle=', ret)
print('threshold_otsu=', ret2)
# threshold= 200.0
# threshold_triangle= 200.0
# threshold_otsu= 117.0

# cup image
src = cv2.imread('./cup.png', 0)
src = cv2.resize(src, dsize=(300,300))
ret, dst = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY)
ret1, dst1 = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY+cv2.THRESH_TRIANGLE)
ret2, dst2 = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)
print('threshold=', ret)
print('threshold_triangle=', ret)
print('threshold_otsu=', ret2)
# threshold= 200.0
# threshold_triangle= 200.0
# threshold_otsu= 106.0

# 결과 비교 
cv2.imshow('original', src)
cv2.imshow('threshold 200', dst)
cv2.imshow('THRESH_TRIANGLE', dst1)
cv2.imshow('THRESH_OTSU', dst2)
cv2.waitKey()
cv2.destroyAllWindows()

 

• cv2.ADAPTIVE_THRESH_MEAN_C
• cv2.ADAPTIVE_THRESH_GAUSSIAN_C

# lena
src = cv2.imread('./lena.jpg', 0)

ret, dst = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY)
ret1, dst1 = cv2.threshold(src, 200, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)
print('global threshold=', ret1)
# global threshold= 117.0

dst2 = cv2.adaptiveThreshold(src, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 51, 0)
dst3 = cv2.adaptiveThreshold(src, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 51, 0)

cv2.imshow('threshold 200', dst)
cv2.imshow('THRESH_OTSU', dst1)
cv2.imshow('ADAPTIVE_THRESH_MEAN_C', dst2)
cv2.imshow('ADAPTIVE_THRESH_GAUSSIAN_C', dst3)
cv2.waitKey()
cv2.destroyAllWindows()

 

# for문 사용해서 최적의 임계값 찾기

import matplotlib.pyplot as plt

src = cv2.imread('./cup.png', 0)

dst_list = []
size_list = []
fig = plt.figure(figsize=(30, 40))
for i in range(21, 182, 20):
    for j in range(21, 102, 10):   
        dst_list.append(cv2.adaptiveThreshold(src, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, i, j))
        size_list.append([i, j])
#        dst_list.append(cv2.adaptiveThreshold(src, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, i, j))
for i in range(1, 82):
    ax = fig.add_subplot(9, 9, i)
    ax.imshow(dst_list[i-1], 'gray')
    ax.set_title(size_list[i-1])
plt.show()

# 4가지 방법 비교
cup1=cv2.imread('./cup.png',0)
ret1,cup1=cv2.threshold(cup1,175,255,cv2.THRESH_BINARY)
ret4,cup4=cv2.threshold(cup,200,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
print('global threshold=',ret4)
# global threshold= 106.0

cup3=cv2.adaptiveThreshold(cup,255,cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY,31,21)
cup4=cv2.adaptiveThreshold(cup,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,31,21)
titles=['binary','binary+otsu','mean','gaussian']
images = [cup1,cup2,cup3,cup4]
for i in range(4):
    plt.subplot(2,2,i+1)
    plt.imshow(images[i],'gray')
    plt.title(titles[i])
    plt.xticks([]),plt.yticks([])
plt.show()