低照度图像恢复算法

基于《Adaptive Local Tone Mapping Based on Retinex for High Dynamic Range Images  》

主要公式如下：

　 是全自适应输出的结果，我们这里就是需要得到他，表示输入图像的luminance值（亮度值），表示输入图像亮度值对的最大值，表示输入亮度对数的平均值，如下式所示

c++主要代码：

void ALTM_Retinex(Mat img)
{
	Mat grayimg;
	cvtColor(img,grayimg,COLOR_RGB2GRAY);
	grayimg.convertTo(grayimg, CV_64FC1, 1.0/255);
	
	int len = grayimg.cols*grayimg.rows;
	double Lwaver, Lwmax;
	Mat gain = grayimg.clone();

	getLwpara(grayimg,len, &Lwaver,&Lwmax);
	
	for (int i = 0; i < grayimg.rows; ++i)
	{
		for (int j = 0; j < grayimg.cols; ++j)
		{
			double tmp = grayimg.at<double>(i, j);
			
			gain.at<double>(i,j) = log(tmp / Lwaver + 1) / log(Lwmax / Lwaver + 1) / tmp;
			if (grayimg.at<double>(i, j) == 0) gain.at<double>(i, j) = 0;
		}
	}
	
	Mat channels[3];
	
	split(img, channels);

	channels[0].convertTo(channels[0], CV_64FC1, 1.0/255);
	channels[1].convertTo(channels[1], CV_64FC1, 1.0/255);
	channels[2].convertTo(channels[2], CV_64FC1, 1.0/255);
	
	channels[0] = gain.mul(channels[0]);
	channels[1] = gain.mul(channels[1]);
	channels[2] = gain.mul(channels[2]);

	Mat finalimg ;
	merge(channels,3,finalimg);

	imshow("final",finalimg);
	waitKey(0);
}

处理效果如图：

参考博客：https://www.cnblogs.com/Imageshop/p/9460334.html