OpenCV 中 parallel_for_理解

OpenCV 中 parallel_for_函数的具体用法，这个函数是OpenCV提供的并行化计算工具，能帮你高效低利用多核CPU加速循环类的计算任务。
一、parallel_for_ 核心概念
parallel_for_ 本质是OpenCV对多线程的封装，它会自动将一个大的循环任务拆分成多个子任务，分配到不同的CPU核心上并行执行，核心优势是：
无需手动管理线程创建/销毁，OpenCV内部自动处理；
适配不同核心数的CPU,兼容性强
语法简洁，只需定义“要并行执行的任务逻辑”即可。
它的核心参数：
/
// void parallel_for_(const Range &range, const ParallelLoopBody &body, double nstripes = -1.);
/
Range:指定循环的范围(如Range(0,1000)对应循环变量从0到999
parallelLoopBody:抽象类，需自定义子类并重写operator()实现具体的并行逻辑；
nstripes:可选，指定拆分的子任务数，-1表示由OpenCV自动决定(推荐)
测试代码：

点击查看代码

#include<opencv2/core.hpp>
#include<opencv2/opencv.hpp>
#include<iostream>
#include<chrono>
#include<filesystem>
#include <windows.h> 
using namespace cv;
using namespace std;
using namespace chrono;
namespace fs = std::filesystem;

//自定义并行任务类：继承 ParallelLoopBody,重写 operator()
class BrightnessAdjustTask :public ParallelLoopBody {
private:
	Mat& img;//待处理图像（引用传递，避免拷贝）
	float brightness; //亮度调整系数
public:
	//构造函数：
	BrightnessAdjustTask(Mat &_img, float _brightness):img(_img),brightness(_brightness){}

	//核心：重写operator(),定义每个线程要执行的子任务
	void operator()(const Range& range) const override {

		//获取当前线程的ID（唯一标识）
		thread::id tid = this_thread::get_id();
		//打印：现成ID+负责的行范围[start,end)
		cout << "[线程" << tid << "] 处理行范围："
			<< range.start << "~" << range.end
			<< "(共" << range.end - range.start << "行) " << endl;
		//range.start -range.end 是当前线程负责的行范围
		for (int i = range.start; i < range.end; i++) {
			//获取第i行的像素指针(uchar 对应 8位图像）
			uchar* row_ptr = img.ptr<uchar>(i);
			for (int j = 0; j < img.cols * img.channels(); j++) {
				//调整亮度，确保像素值在0~255范围内
				row_ptr[j] = saturate_cast<uchar>(row_ptr[j] * brightness); //see1 saturate_cast
			}
		}
	}

};

int main() {
	//1.读取图像
	//Mat img = imread("D:\\pic\\shalabowa.jpeg",1);
	string img_path = "E:\\VSWorkSpace\\CppLearn\\CppLearn\\x64\\Debug\\1.jpg";

	Mat img = cv::imread(img_path, IMREAD_COLOR);

	cout << "读取成功！图片尺寸：" << img.rows << "x" << img.cols << endl;
	//imshow("测试图片", img);
	//waitKey(0); // 按任意键关闭窗口

	if (img.empty()) {
		cout << "图像读取失败！" << endl;
		return -1;
	}
	float brightness = 1.5;// 亮度提高50%

	//2. 串行遍历（对比性能）
	auto start_serial = high_resolution_clock::now(); //see 怎么理解high_resolution_clock
	Mat img_serial = img.clone();
	for (int i = 0; i < img_serial.rows; i++)
	{
		uchar* row_ptr = img_serial.ptr<uchar>(i);
		for (int j = 0; j < img_serial.cols * img_serial.channels(); j++)
		{
			row_ptr[j] = saturate_cast<uchar>(row_ptr[j] * brightness);
		}
	}

	auto end_serial = high_resolution_clock::now();
	double time_serial = duration_cast<milliseconds>(end_serial - start_serial).count();
	cout << "串行执行时间：" << time_serial << " ms" << endl;
   
	//3.并行遍历（使用parallel_for_)
	auto start_parallel = high_resolution_clock::now();
	Mat img_parallel = img.clone();
	//定义并行任务对象
	BrightnessAdjustTask task(img_parallel, brightness);
	//执行并行循环:Range(0,img.rows)表示按行拆分任务
	parallel_for_(Range(0, img_parallel.rows), task,5);
	auto end_parallel = high_resolution_clock::now();
	double time_parallel = duration_cast<milliseconds>(end_parallel - start_parallel).count();
	cout << "并行执行时间：" << time_parallel << " ms" << endl;

	imwrite("D:\\pic\\see\\parallel_test\\serial_result.jpg", img_serial);
	imwrite("D:\\pic\\see\\parallel_test\\parallel_result.jpg", img_parallel);

}

测试结果如下：可以看出即使在多线程中增加了一些打印语句速度也明显提升

点击查看代码

读取成功！图片尺寸：1600x2406
串行执行时间：255 ms
[ INFO:0@0.499] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\parallel\registry_parallel.impl.hpp (96) cv::parallel::ParallelBackendRegistry::ParallelBackendRegistry core(parallel): Enabled backends(3, sorted by priority): ONETBB(1000); TBB(990); OPENMP(980)
[ INFO:0@0.501] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load E:\opencv\build\x64\vc15\bin\opencv_core_parallel_onetbb460_64d.dll => FAILED
[ INFO:0@0.503] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load opencv_core_parallel_onetbb460_64d.dll => FAILED
[ INFO:0@0.503] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load E:\opencv\build\x64\vc15\bin\opencv_core_parallel_tbb460_64d.dll => FAILED
[ INFO:0@0.505] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load opencv_core_parallel_tbb460_64d.dll => FAILED
[ INFO:0@0.506] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load E:\opencv\build\x64\vc15\bin\opencv_core_parallel_openmp460_64d.dll => FAILED
[ INFO:0@0.507] global c:\build\master_winpack-build-win64-vc15\opencv\modules\core\src\utils\plugin_loader.impl.hpp (67) cv::plugin::impl::DynamicLib::libraryLoad load opencv_core_parallel_openmp460_64d.dll => FAILED
[线程23268] 处理行范围：1280~1600(共320行)
[线程20792] 处理行范围：0~320(共320行)
[线程23868] 处理行范围：320~640(共320行)
[线程23256] 处理行范围：960~1280(共320行)
[线程27300] 处理行范围：640~960(共320行)
并行执行时间：112 ms

E:\VSWorkSpace\CppLearn\x64\Debug\CppLearn.exe (进程 22056)已退出，代码为 0。
要在调试停止时自动关闭控制台，请启用“工具”->“选项”->“调试”->“调试停止时自动关闭控制台”。