问题:

I have a byte array that represents a .jpg file that I want to convert directly to an OpenCV Mat object.

I have something like

byte* data; // Represents a JPG that I don't want to disk and then read.
// What goes here to end up with the following line?
cv::Mat* image_representing_the_data;

回答1:

the previously mentioned method will work fine, if it's PIXEL data.

if instead, you have a whole jpg FILE in memory, headers, compression, and all, it won't work.

in that case you want:

Mat img = imdecode(data);

which will do the same as imread(), only from memory, not from a filename

回答2:

as @bob mention, this can be done using the Mat constructor.

byte *data;
cv::Mat imageWithData = cv::Mat(sizeOfData, 1, CV_8U, data).clone();

After you have created this matrix, call the reshape function with the number of rows in the image.

cv::Mat reshapedImage = imageWithData.reshape(0, numberOfRows);

回答3:

cv::Mat GetImageFromMemory(uchar* image, int length, int flag)
    {
        std::vector<uchar> data = std::vector<uchar>(image, image + length);
        cv::Mat ImMat = imdecode(data, flag);

        return ImMat;
    }

the parameter "flag" refers to the type of image, grayscale, color, any depht .. that has opencv defined