cereal进行数据序列化

cv::Mat

cereal不直接支持cv::Mat类型的数据结构，但是可以通过自己设计的结构体，进行序列化数据存储和读取。

#include <iostream>
#include <opencv2/opencv.hpp>
#include <cereal/archives/binary.hpp>
#include <cereal/types/vector.hpp> 
#include <fstream>


struct imgInfo{
    int32_t imgWidth;
    int32_t imgHeight;
    int32_t imgChannel;
    int32_t imgType;
    std::vector<uint8_t> imgData;
	//结构体中必须包含serialize函数才能使用cereal进行序列化
    template<class Archive>
    void serialize(Archive & archive){
    	archive(imgWidth, imgHeight, imgChannel, imgType, imgData);
    }
};

void main(){
    cv::Mat testImg = cv::imread("/Users/Downloads/test.jpeg");

    testInfo outInfo;
    outInfo.imgWidth = testImg.cols;
    outInfo.imgHeight = testImg.rows;
    outInfo.imgChannel = testImg.channels();
    outInfo.imgType = testImg.type();

    if (testImg.isContinuous()) {
        outInfo.imgData.assign(testImg.data, testImg.data + testImg.total()*testImg.channels());
    } else {
        for (int i = 0; i < testImg.rows; ++i){
            outInfo.imgData.insert(outInfo.imgData.end(), testImg.ptr<uchar>(i), testImg.ptr<uchar>(i)+testImg.cols*testImg.channels());
        }
    }
	//存储二进制图像数据
    {
        std::ofstream os("out.lut", std::ios::out | std::ios::binary);
        cereal::BinaryOutputArchive archive(os);
        archive(outInfo); 
    }
    //读取二进制图像数据
    testInfo inInfo;
    {
        std::ifstream intest("out.lut", std::ios::in | std::ios::binary);
        cereal::BinaryInputArchive iarchive(intest);
        iarchive(inInfo);
    }

    cv::Mat inImg= cv::Mat{inInfo.imgWidth, inInfo.imgHeight, inInfo.imgType,inInfo.imgData.data()};
}

Tips:

1. mat.isContinuous返回矩阵元素在内存中是否连续存储。如果图像数据在每行末尾是没有间隙连续存储的，则方法返回true。一般经过裁剪或截取的mat图像不是连续存储的。
2. 出现错误：error: static_assert failed due to requirement 'traits::detail::count_output_serializers<std::__1::vector<unsigned char, std::__1::allocator<unsigned char> >, cereal::BinaryOutputArchive>::value != 0' "cereal could not find any output serialization functions for the provided type and archive combination. \n\n Types must either have a serialize function, load/save pair, or load_minimal/save_minimal pair (you may not mix these).
   解决方法：添加#include <cereal/types/vector.hpp>。序列化vector时需要添加相应的头文件。序列化其他数据类型时，需要添加其他相应的头文件。

参考资料

1. https://uscilab.github.io/cereal/quickstart.html