来自表示RGB图像的浮点数组的初始CV :: MAT

Question

Introduction to my problem:
I have serialized myself a CV::MAT object in c++ and sent in to a python program (more accurately, I have a python program that uses a c++ .exe as subprocess..), I deserialize it in python back into a 2D array of (R,G,B) ints

Now I have a different method in my python service that on demand sends this 2D array back to the c++ process and have it do something I DONT want to use boost serializer because its basically hell I have serialized it myself from python and manage to give the c++ process a flattened array along with the dimentions of the 2D array (rows and cols)

Now I have no idea how I construct a CV::MAT object back from this data i know there's Mat (int rows, int cols, int type, void *data, size_t step=AUTO_STEP) which might do the job but no idea what goes insize the step in case its what I think it is

Any help would be highly appreciated!

This is the c++ code I have so far

    vector<string> imgVector;
    int imgRows = stoi(kp.at(7));
    int imgCols = stoi(kp.at(8));
    split(kp.at(2), ' ', back_inserter(imgVector)); // kp.at(2) is a ' ' seperated array of integers
    int* pixels = (int*)malloc(sizeof(int) * imgVector.size());
    for (int j = 0; (unsigned)j < imgVector.size(); i++){
        pixels[j] = stoi(imgVector[j]);
    }
    Mat image(imgRows, imgCols, CV_32F, pixels);

But I think it will make a 2D array of integers and not of RGB tuples like I want

Answer 1

The corrected items:

• You need the type to be CV_8UC3 to store a 3-channel int array.
• The size of the mat should be rows*cols*3.
• The data type per pixel should be uchar .

Following those, the following code should work.

vector<string> imgVector;
int imgRows = stoi(kp.at(7));
int imgCols = stoi(kp.at(8));
split(kp.at(2), ' ', back_inserter(imgVector)); // kp.at(2) is a ' ' seperated array of integers
uchar* pixels = (uchar*)malloc(sizeof(uchar) * imgRows*imgCols*3);
for (int j = 0; (unsigned)j < imgVector.size(); i++){
    pixels[j] = uchar(stoi(imgVector[j]));
}
Mat image(imgRows, imgCols, CV_8UC3, pixels);