颜色直方图实验

学习颜色空间和颜色直方图，使用OpenCV + Python进行一些小实验。

实验介绍

对图片进行颜色空间的转换

画出图片的颜色直方图

对两张图片的颜色直方图进行比较

实验环境

操作系统：Ubuntu 14.04.3 LTS

（刚开始用Windows 10，然后发现用Python的PIL读取jpg文件时得到的RGB编码与Ubuntu下不同，而bmp文件却是一致的。经测试在Ubuntu下用Python得到jpg文件的RGB编码与mspaint一致，故改用Ubuntu。）

开发环境：Python 2.7.6 + OpenCV 2.4.11

（OpenCV 3.x与OpenCV 2.x略有不同。）

Python Library:

numpy 1.10.2

matplotlib 1.5.0

Pillow 3.1.1

实验过程

查看不同颜色空间的编码

输入：图片文件路径、颜色空间

输出：编码

参考：http://docs.opencv.org/3.0-beta/doc/py_tutorials/py_imgproc/py_colorspaces/py_colorspaces.html#converting-colorspaces

读入图片（OpenCV中默认颜色空间为BGR）

转换颜色空间

输出每个通道的编码

import cv2

input_filepath = 'Lenna.png'
output_filepath = ['R.txt', 'G.txt', 'B.txt']
img = cv2.imread(input_filepath)
imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# imgHSV = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# imgLAB = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
for k in range(3):
fd = open(output_filepath[k], 'w')
for i in imgRGB[:, :, k]:
fd.write(' '.join(['{:3}'.format(j) for j in i]) + '\n')
fd.close()

画出颜色直方图

输入：图片文件路径、颜色空间

输出：颜色直方图

参考：http://docs.opencv.org/3.0-beta/doc/py_tutorials/py_imgproc/py_histograms/py_table_of_contents_histograms/py_table_of_contents_histograms.html#table-of-content-histograms

http://docs.opencv.org/3.0-beta/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.html#histogram-calculation

读入图片

转换颜色空间

获得颜色直方图（可量化，注意HSV的H通道的大小是180）

绘制颜色直方图

import cv2
import numpy as np
import matplotlib.pyplot as plt

input_filepath = 'Lenna.png'
img = cv2.imread(input_filepath)
# RGB
imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
hist = [cv2.calcHist([imgRGB], [k], None, [256], [0, 256]) for k in range(3)]
x = np.arange(256) + 0.5
plt.subplot(221), plt.imshow(imgRGB)
plt.subplot(222), plt.bar(x, hist[0], color = 'r', edgecolor = 'r')
plt.subplot(223), plt.bar(x, hist[1], color = 'g', edgecolor = 'g')
plt.subplot(224), plt.bar(x, hist[2], color = 'b', edgecolor = 'b')
plt.show()
# HSV
imgHSV = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
hist = [cv2.calcHist([imgHSV], [0], None, [50], [0, 180]), \
cv2.calcHist([imgHSV], [1], None, [50], [0, 256])]
x = np.arange(50) + 0.5
plt.subplot(211), plt.bar(x, hist[0])
plt.subplot(212), plt.bar(x, hist[1])
plt.show()

比较颜色直方图

输入：两张图片文件路径、颜色空间、比较方法

输出：比较结果（一个实数值）

参考：http://docs.opencv.org/3.0-beta/doc/tutorials/imgproc/histograms/histogram_comparison/histogram_comparison.html#histogram-comparison

读入图片

转换颜色空间

获得颜色直方图并归一化

比较颜色直方图

import cv2
import numpy as np
import matplotlib.pyplot as plt

input_filepath = ['1.jpg', '2.jpg']
comp_method = [cv2.cv.CV_COMP_CORREL, cv2.cv.CV_COMP_INTERSECT, \
cv2.cv.CV_COMP_CHISQR, cv2.cv.CV_COMP_BHATTACHARYYA]
# BGR
img = [cv2.imread(i) for i in input_filepath]
hist = [cv2.calcHist([i], [k], None, [256], [0, 256]) for k in range(3) for i in img]
for i in hist
for j in i:
cv2.normalize(j, j)
hist = [np.mean(i, 0) for i in hist]
for method in comp_method:
d = cv2.compareHist(hist[0], hist[1], method)
print(d)
# HSV
imgHSV = [cv2.cvtColor(i, cv2.COLOR_BGR2HSV) for i in img]
hist = [cv2.calcHist([i], [0, 1], None, [50, 50], [0, 180, 0, 256]) \
for i in imgHSV]
for i in hist_set:
cv2.normalize(i, i)
for method in comp_method:
d = cv2.compareHist(hist[0], hist[1], method)
print(d)

参考

直方图. https://zh.wikipedia.org/wiki/%E7%9B%B4%E6%96%B9%E5%9B%BE#.E9.A2.9C.E8.89.B2.E7.9B.B4.E6.96.B9.E5.9B.BE

颜色直方图. http://baike.baidu.com/view/2438797.htm

OpenCV-Python Tutorials. http://docs.opencv.org/3.0-beta/doc/py_tutorials/py_tutorials.html

Lenna. https://upload.wikimedia.org/wikipedia/en/2/24/Lenna.png