OpenCV 统计学 —— 直方均衡补偿

#include<opencv2/opencv.hpp>
#include<iostream>

using namespace std;
using namespace cv;

// 计算 mat 的直方图的统计
void calHist(const Mat &mat, Mat &hist) {
	// 初始化 hist
	hist.create(1, 256, CV_32S);
	for (int i = 0; i < hist.cols; i++)
	{
		hist.at<int>(0, i) = 0;
	}
	// 统计 mat 中, 每一个像素值出现的次数
	for (int row = 0; row < mat.rows; row++) {
		for (int col = 0; col < mat.cols; col++) {
			int index = mat.at<uchar>(row, col);
			hist.at<int>(0, index)++;
		}
	}
}

// 将 src 中的数据归一化到 [0, edge] 中
void normalize(const Mat &src, Mat dst, int edge) {
	// 1. 统计最大值
	float max_value = 0;
	for (int row = 0; row < src.rows; row++)
	{
		for (int col = 0; col < src.cols; col++) {
			int val = src.at<int>(row, col);
			if (val > max_value) {
				max_value = val;
			}
		}
	}
	cout << max_value << endl;
	// 2. 计算最大值与边界的比率
	float ratio = edge / max_value;
	// 3. 将 src 中的数据, 归一化到 [0, edge] 中
	dst.create(src.size(), src.type());
	for (int row = 0; row < src.rows; row++)
	{
		for (int col = 0; col < src.cols; col++) {
			dst.at<int>(row, col) = src.at<int>(row, col) * ratio;
		}
	}
}

// 绘制直方图
void drawHist(int draw_width, int draw_height, Mat hist) {
	Mat hist_image(draw_height, draw_width, CV_8SC4, Scalar(255, 255, 255));
	int paintWidth = draw_width / hist.cols;
	for (int i = 1; i < hist.cols; i++) {
		line(
			hist_image,
			Point((i - 1) * paintWidth, draw_height - hist.at<int>(i - 1)),
			Point(i * paintWidth, draw_height - hist.at<int>(i)),
			Scalar(255, 0, 0),
			paintWidth,
			LINE_AA
		);
	}
	imshow("hist", hist_image);

}

// 1. 直方图的统计
// 2. 计算直方图中像素的概率
// 3. 生成映射表
// 4. 从映射表中查找数据
void equalizeHist(const Mat &src, Mat &dst) {
	Mat hist;
	// 1. 统计直方图 -> hist
	calHist(src, hist);
	// 2. 计算直方图中像素的累加概率
	Mat prob_mat(hist.size(), CV_32FC1);
	float total_pixels = src.cols * src.rows;
	for (int i = 0; i < hist.cols; i++) {
	    float prob = hist.at<int>(0, i) / total_pixels;
		prob_mat.at<float>(0, i) = prob;
	}
	// 3. 计算累加概率
	float prob_sum = 0;// 累加概率
	for (int i = 0; i < hist.cols; i++) {
		prob_sum += prob_mat.at<float>(0, i);
		prob_mat.at<float>(0, i) = prob_sum;
	}
	// 4. 通过概率生成像素值的直方映射表
	Mat map(hist.size(), hist.type());
	for (int i = 0; i < hist.cols; i++) {
		float prob = prob_mat.at<float>(0, i);
		map.at<int>(0, i) = prob * 255; 
	}
	// 5. 从映射表中查找数据
	dst.create(src.size(), src.type());
	for (int row = 0; row < src.rows; row++) {
		for (int col = 0; col < src.cols; col++) {
			int pixel = src.at<uchar>(row, col);
			int map_pixel = map.at<int>(0, pixel);
			dst.at<uchar>(row, col) = map_pixel;
		}
	}
}

void main() {
	Mat src = imread("F:/VisualStudioSpace/OpenCV/src/Naruto.jpg");
	// imshow("src", src);
	// 获取灰度图
	Mat gray;
	cvtColor(src, gray, COLOR_BGR2GRAY);
	imshow("gray", gray);
	// 对灰度图进行直方均衡
	Mat equalize_mat;
	equalizeHist(gray, equalize_mat);
	imshow("equalize_mat", equalize_mat);
	cvWaitKey(0);
}

#include<opencv2/opencv.hpp>

using namespace cv;

void main() {
	Mat src = imread("F:/VisualStudioSpace/OpenCV/Resource/Naruto.jpg");
	if (!src.data) {
		printf("Read data failed");
		return;
	}
	imshow("src", src);
	// 转灰度
	Mat gray;
	cvtColor(src, gray, COLOR_BGR2GRAY);
	imshow("gray", gray);
	// 直方均衡补偿
	Mat dst;
	equalizeHist(gray, dst);
	imshow("equalize", dst);
	cvWaitKey(0);
}

#include<opencv2/opencv.hpp>
#include<vector>

using namespace cv;
using namespace std;

void main() {
	Mat src = imread("F:/VisualStudioSpace/OpenCV/Resource/equalize.jpg");
	if (!src.data) {
		printf("Read data failed");
		return;
	}
	imshow("src", src);
	// 1. 转 HSV
	Mat hsv;
	cvtColor(src, hsv, COLOR_BGR2HSV);
	// 2. 提取 V 分量
	vector<Mat> hsvs;
	split(hsv, hsvs);
	// 3. 对 V 分量进行直方均衡补偿
	Mat equalize_v;
	equalizeHist(hsvs[2], equalize_v);
	// 4. 合入原图
	// merge(equalize_v, hsv)
	for (int i = 0; i < hsv.rows; i++)
	{
		for (int j = 0; j < hsv.cols; j++) {
			hsv.at<Vec3b>(i, j)[2] = equalize_v.at<uchar>(i, j);
		}
	}
	// 5. 转 BGR
	Mat dst;
	cvtColor(hsv, dst, COLOR_HSV2BGR);
	imshow("equalized", dst);

	cvWaitKey(0);
}