opencv-contrib-python 的ColorCorrectionModel模块进行色彩校正

1.安装opencv-contrib-python，如果之前安装了opencv-python要先卸载

python -m pip install opencv-contrib-python

2.代码如下：

import cv2
import numpy as np

import os

image_path = 'E:/datasets/biseka.png'

img = cv2.imdecode(np.fromfile(image_path,dtype=np.uint8),-1)

cv2.imshow('img', img)
cv2.waitKey(0)

detector = cv2.mcc.CCheckerDetector_create()

detector.process(img, cv2.mcc.MCC24)

# cv2.mcc_CCheckerDetector.getBestColorChecker()
checker = detector.getBestColorChecker()

cdraw = cv2.mcc.CCheckerDraw_create(checker)
img_draw = img.copy()
cdraw.draw(img_draw)

cv2.imshow('img_draw', img_draw)
cv2.waitKey(0)

chartsRGB = checker.getChartsRGB()

src = chartsRGB[:,1].copy().reshape(24, 1, 3)

src /= 255.0

print(src.shape)
# model1 = cv2.ccm_ColorCorrectionModel(src, cv2.mcc.MCC24)
model1 = cv2.ccm_ColorCorrectionModel(src, cv2.ccm.COLORCHECKER_Macbeth)
# model1 = cv2.ccm_ColorCorrectionModel(src,src,cv2.ccm.COLOR_SPACE_sRGB)
model1.run()
ccm = model1.getCCM()
print("ccm ", ccm)
loss = model1.getLoss()
print("loss ", loss)

img_ = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_ = img_.astype(np.float64)
img_ = img_/255
calibratedImage = model1.infer(img_)
out_ = calibratedImage * 255
out_[out_ < 0] = 0
out_[out_ > 255] = 255
out_ = out_.astype(np.uint8)

out_img = cv2.cvtColor(out_, cv2.COLOR_RGB2BGR)

file, ext = os.path.splitext(image_path)
calibratedFilePath = file + '_calibrated' + ext
# cv2.imwrite(calibratedFilePath, out_img)
cv2.imencode(ext, out_img)[1].tofile(calibratedFilePath)

cv2.imshow('out_img', out_img)
cv2.waitKey(0)

 

参考文献：

1.https://docs.opencv.org/master/index.html

2.https://docs.opencv.org/master/de/df4/group__color__correction.html

3.https://docs.opencv.org/master/dd/d19/group__mcc.html

4.https://docs.opencv.org/master/d2/dee/classcv_1_1mcc_1_1CCheckerDraw.html

5.https://docs.opencv.org/master/db/d3e/classcv_1_1mcc_1_1CChecker.html