cv::resize function costs 143M GPU memory in opencv version 3.2.0

edit

Below is my test script:

#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/highgui/highgui.hpp>
//#include <opencv2/imgcodecs/imgcodecs.hpp>

#include <iostream>
#include <string>

int main() {
    std::string path = "/home/pangwong/TestSpace/test_images/car_test.png";
    cv::Mat img = cv::imread(path);
    for (int i = 0; i < 100000; i++) {
        std::cin.get();
        cv::resize(img, img, cv::Size(144, 144));
        std::cout << "aaa" << std::endl;
        std::cin.get();
    }
}

When the routine went to the first "std::cin.get()", no memory was occupied. But when I input something and went to the second "std::cin.get()", something strange happened. Bolow is the nvidia-smi result.

As we can see, 143M GPU memory has been occupied.

System Info:

1. Ubuntu 16.04
2. Opencv3.2.0, cmake compile flag WITH_CUDA=OFF
3. Nvidia GTX1080Ti, CUDA8.0

Anybody could explain what happened and how can I release the memory consumption?

Thanks in advance.