Find QImage inside another QImage.

Synny

I have two QImage objects. I want to find the x,y of the smaller Qimage in the larger QImage.
Update: The smaller image is a perfect crop of the larger image. Just to find a sub matrix in a matrix of rgb.
How can I do it?

mrjj

Hi and welcome to the forums
So image 2 is also draw somewhere on image 1 ?

It depends on what is around the image 2 when its inside image 1
if a solid color or not very noise you might be able to search for a bit pattern and
get no false positives.
If you need something more advanced, then something like
https://docs.opencv.org/2.4/doc/tutorials/imgproc/histograms/template_matching/template_matching.html

Synny

The smaller image is a perfect crop of the larger image. No need to draw a thing. No need for complicated match algorithms. Just to find a sub matrix in a matrix of rgb.

SGaist

Hi and welcome to devnet,

Then you'll have to implement your own function that will find the first pixel, then ensure that the second is the right etc.

Pablo J. Rogina

@Synny maybe template matching using OpenCV can help you. Although some of the examples you can find are Python-based, I guess those could be a good start point to implement your C++ code.

Synny

I wrote a function. Works great :)

QPoint FindImageInImage(QImage &ImgLarge , QImage &ImgSmall )
{
    if ( ImgLarge.format() != QImage::Format_RGB32 )
        ImgLarge.convertToFormat(QImage::Format_RGB32);

    if ( ImgSmall.format() != QImage::Format_RGB32 )
        ImgSmall.convertToFormat(QImage::Format_RGB32);

    const uchar* bitsImgLarge = ImgLarge.constBits();
    int nHeightImgLarge = ImgLarge.height();
    int nWidthImgLarge = ImgLarge.width();

    const uchar* bitsImgSmall = ImgSmall.constBits();
    int nHeightImgSmall = ImgSmall.height();
    int nWidthImgSmall = ImgSmall.width();

    QPoint ptFound = QPoint(-1,-1);

    for ( int yInImgLarge = 0 ; yInImgLarge < nHeightImgLarge-nHeightImgSmall ; yInImgLarge++  )
        for ( int xInImgLarge = 0 ; xInImgLarge < nWidthImgLarge-nWidthImgSmall ; xInImgLarge++  )
        {
            bool bFound = true;

            for ( int yInImgSmall = 0 ; yInImgSmall < nHeightImgSmall ; yInImgSmall++  )
                for ( int xInImgSmall = 0 ; xInImgSmall < nWidthImgSmall ; xInImgSmall++  )
                {
                    int nXInImgLarge_ = xInImgLarge+xInImgSmall;
                    int nYInImgLarge_ = yInImgLarge+yInImgSmall;

                    int nXInImgSmall_ = xInImgSmall;
                    int nYInImgSmall_ = yInImgSmall;

                    if ( nYInImgLarge_ >= nHeightImgLarge )
                    {
                        bFound = false;
                        goto endInnerLoop;
                    }

                    if ( nXInImgLarge_ >= nWidthImgLarge )
                    {
                        bFound = false;
                        goto endInnerLoop;
                    }

                    const uchar* pImgLargeCurr = bitsImgLarge + 4*((nYInImgLarge_)*nWidthImgLarge+nXInImgLarge_);
                    const uchar* pImgSmallCurr = bitsImgSmall + 4*(nYInImgSmall_*nWidthImgSmall+nXInImgSmall_);

                    if ( memcmp( (void*)pImgLargeCurr, (void*)pImgSmallCurr, 3 ) )
                    {
                        bFound = false;
                        goto endInnerLoop;
                    }
                }

            endInnerLoop:

            if ( bFound )
            {
                ptFound = QPoint( xInImgLarge , yInImgLarge );
                goto endOuterLoop;
            }
        }

    endOuterLoop:

    return ptFound;
}

mrjj

Hi
Thank you for sharing.
Why the goto if i may ask, and not just return ptFound;
in case of clean up or ?