Question

I am new to OpenCV. Currently, trying to load and save a defined ROI of an image.
For OpenCV 1.x, I got it working with the following function…

#include <cv.h>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>  

void SaveROI(const CStringA& inputFile, const CStringA& outputFile)
{
    if (ATLPath::FileExists(inputFile))
    {
        CvRect rect;
        rect.x      = 8;
        rect.y      = 90;
        rect.width  = 26;
        rect.height = 46;

        IplImage* imgInput = cvLoadImage(inputFile.GetString(), 1); 
        IplImage* imgRoi   = cvCloneImage(imgInput);
        cvSetImageROI(imgRoi, rect);
        cvSaveImage(outputFile.GetString(), imgRoi);

        cvReleaseImage(&imgInput);
        cvReleaseImage(&imgRoi);
    }
}

How can this be done with the OpenCV 2 or C++. I tried the following without a success, the whole image is saved.

void SaveROICPP(const CStringA& inputFile, const CStringA& outputFile)
{      
    if (ATLPath::FileExists(inputFile))
    {
        cv::Mat imgInput = cv::imread(inputFile.GetString());

        if (imgInput.data != NULL)
        {
            cv::Mat imgRoi = imgInput(cv::Rect(8, 90, 26, 46));

            imgInput.copyTo(imgRoi);                 

            cv::imwrite(outputFile.GetString(), imgRoi);
        }
    }
}

Any help or suggestion?

Answer 1

You just don’t need to call copyTo:

void SaveROICPP(const CStringA& inputFile, const CStringA& outputFile)
{      
    if (ATLPath::FileExists(inputFile))
    {
        cv::Mat imgInput = cv::imread(inputFile.GetString());

        if (imgInput.data != NULL)
        {
            cv::Mat imgRoi = imgInput(cv::Rect(8, 90, 26, 46));
            cv::imwrite(outputFile.GetString(), imgRoi);
        }
    }
}

In your version copyTo sees that imgInput is bigger then imgRoi and reallocates a new full-size matrix to make the copy. imgRoi is already a sub-image and you can simply pass it to any OpenCV function.