为图像对比度指标实施直方图扩展

Question

目前我正在尝试根据本文计算图像对比度指标的直方图分布 Performance metrics for image contrast. The algorithm for easier and faster to read showed by this screenshot of the paper: 或者引用论文：

Histogram spread is the ratio of the quartile distance to the
range of the histogram. Quartile distance is denoted as the
difference between the 3 rd quartile and the 1 st quartile. Here 3 rd
quartile and 1 st quartile means the histogram bins at which
cumulative histogram have 75% and 25% of the maximum
value respectively (see Fig.1)

我在这个实验中使用了这两张图像，它们都是医学图像（乳房 X 线照片）
1.

2.

这是我使用的代码：

import cv2
import numpy as np


def calculate_histogram_spread(gray_img):

    #compute image histogram and its cdf
    histo,_ = np.histogram(gray_img.flatten(),256,[0,256])
    cdf = np.cumsum(histo)

    #calculate the Q3 (75%percentile) and Q1 (25%percentile) 
    Q3 = np.percentile(cdf,75)
    Q1 = np.percentile(cdf,25)

    #find the which bin value  
    Q3_bin = np.searchsorted(cdf,Q3) 
    Q1_bin = np.searchsorted(cdf,Q1)
    
    IQR = Q3_bin - Q1_bin
    divider = np.max(gray_img) - np.min(gray_img)

    HS = IQR/divider
    return Q3_bin,Q1_bin,HS

if __name__ == '__main__':

    sample1 = '/home/mario/Desktop/hs1.jpg'
    sample2 = '/home/mario/Desktop/hs2.jpg'
 
    img1 = cv2.imread(sample1,0)
    img2 = cv2.imread(sample2,0)

    print('Sample1')
    Q3_bin,Q1_bin,HS = calculate_histogram_spread(img1)
    print('Q3_bin={},Q1_bin={},HS={}'.format(Q3_bin,Q1_bin,HS))

    print('Sample2')
    Q3_bin,Q1_bin,HS = calculate_histogram_spread(img2)
    print('Q3_bin={},Q1_bin={},HS={}'.format(Q3_bin,Q1_bin,HS))

这是我的最终结果：

Sample1
Q3_bin=192,Q1_bin=64,HS=0.5019607843137255
Sample2
Q3_bin=192,Q1_bin=64,HS=0.5019607843137255

我的问题是，为什么两张图片的输出完全相同，而在视觉上图像的对比度不同（我认为），我做错了吗？有谁能够帮助我。我还在以相同的尺寸/分辨率测试这些图像。

Answer 1

这似乎在 Python/OpenCV 中为我提供了不同的对比度值。

输入 1:

输入 2：

import cv2
import numpy as np

# load images as grayscale
img = cv2.imread("breast1.jpg", 0)
#img = cv2.imread("breast2.jpg", 0)
hh, ww = img.shape[:2]

# compute total pixels
tot = hh * ww

# compute histogram
hist = np.histogram(img,bins=256,range=[0,255])[0]

# compute cumulative histogram
cum = np.cumsum(hist)

# normalize histogram to range 0 to 100
cum = 100 * cum / tot

# get bins of percentile at 25 and 75 percent in cum histogram
i = 0
while cum[i] < 25:
    i = i+1
B1 = i
i = 0
while cum[i] < 75:
    i = i+1
B3 = i
print('25 and 75 percentile bins:',B1,B3)

# compute min and max graylevel (which are also the min and max bins)
min = np.amin(img)
max = np.amax(img)
print('min:',min,  'max:',max)

# compute contrast
contrast = (B3-B1)/(max-min)
print('contrast:',contrast)

图片 1 的结果：

25 and 75 percentile bins: 0 147
min: 0 max: 255
contrast: 0.5764705882352941

图片 2 的结果：

25 and 75 percentile bins: 58 162
min: 0 max: 255
contrast: 0.40784313725490196

为图像对比度指标实施直方图扩展

Implementing Histogram Spread for Image Contrast Metrics

python

opencv

image

histogram