Accessing pixel pointer

edit

You should have a look at that page for an overview of pixels' access in OpenCV.

But for a short answer, you could do like this:

float* p;
float* q;
for( int i = 0; i < image.rows; ++i)
{
    p = image.ptr< float >( i );
    q = image2.ptr< float >( i ); // I guess image2 is the same size as image1
    // and both image are in float!
    for ( int j = 0; j < image.cols; ++j)
    {
        // Do whatever you want
        p[ j ] = 255.f * q[ j ] / 3.5f;
    }
}

ROS image is an image in the Robot Operating System. cv_ptr is a pointer to that image. For a faster access a float pointer is made (I would prefer to use cv::Ptr, std::unique_ptr, or std::auto, instead). Every image.ptr<float>(i) is a pointer to the i-th row of the matrix image. Every element of that row pointer is an element on a column (row i, col j). So by the pointer, the element is accessed; the pointer points to the address of that float (image pixel).

I am not sure what exactly ptr_img_B is doing, but some kind of operation on that pixel (because ptr_img_A is incremented)