Solved: Magick++ Version of 3 image masked alpha composition

[The_Chaos]

Hi everybody,

currently I am trying to figure out how to write a c++ program using the magick++ libs to recreate the behaviour for the "special three image form of Masked Alpha Composition" mentioned in http://www.imagemagick.org/Usage/channe ... ed_compose.

As far as I understood, parts of the idea is by using the mask image with both the Multiply composite flag (like in image1.composite( image2, placement, MultiplyCompositeOp );) and the Screen composite flag.

Although I found the MultiplyCompositeOp define for magick++, I couldn't find the 'convert -composite Screen' equivalent.
Did I miss something here or is this method not available?

Has anyone a piece of c++ code that actually does the same like the 3 image form from above?

Any help or hint would be greatly appreciated.
Olli

[el_supremo]

I have a MagickWand example in C of how to do that but it involves resorting to MagickCore to set an image mask.
http://members.shaw.ca/el.supremo/Magic ... ipmask.htm

Pete

[The_Chaos]

Hi el_supremo,

thank you very much for your help and hints! This is exactly what I needed!

Just for future references and in case somebody wants or needs the same thing, here is my piece of code that works:

Code: Select all

#include <Magick++.h>
#include <iostream>
using namespace std;
using namespace Magick;
int main(int argc,char **argv)
{
	// create image objects for use
	Image src, dest, mask;

	try {
		// read image files into objects
		dest.read( "imageWithPlaceToRenderPhotoIn.png" );
		mask.read( "imageWithMaskForPhotoPlace.png" );
		src.read( "photoToRenderIn.jpg" );

		// create parameters for image distortion of the src image
		double distortParams[] = {0,0, 82.5,554.5, 0,1200, 532.5,764.5, 800,1200, 786.5,256.5, 800,0, 422.5,118.5};

		// set surrounding pixels to transparent
		src.virtualPixelMethod(TransparentVirtualPixelMethod);

		// distort the src image to fit inside the masked area
		src.distort(MagickCore::PerspectiveDistortion, 16, distortParams, 0);

		// negate the mask, since we need black pixels where content has to be seen
		// although the mask is made with white pixels where content shines through
		mask.negate();

		// combine the destination image and the mask at the origin
		Geometry placement(0,0,0,0);
		dest.clipMask(mask);
		dest.composite( src, placement, OverCompositeOp);

		// Write the destination image to a file
		dest.write( "finalImageWithPhotoRenderedIn.png" );
	}
	catch( Exception &error_ )
	{
		cout << "Caught exception: " << error_.what() << endl;
		return 1;
	}
	return 0;
}

Once again, thank you very much for your help!

Greetings,
Olli

[el_supremo]

And thank you for posting your code. I hadn't realized that MagickSetImageClipMask would work in that example. I had been looking for something that set wand->images->mask (which is what the convert/mogrify code does) but it seems that using MagickSetImageClipMask to set wand->images->clip_mask will still produce the desired result. I have updated the clipmask example.

Pete

[The_Chaos]

Hi Pete,

And thank you for posting your code.

Well thats the least I could do

But I still got another question regarding the command line tools and the magick++ equivalent:
When I try to put an Overlay over an image to create shadows/highlights, I can use the following command:

Code: Select all

composite image.png overlay.png -compose Overlay final.png

which uses a grayscale overlay image and darkens the part in the source image, where the overlay is dark and lightens the source where the overlay is bright. It looks like every grayscale value >127 darkens and <127 lightens the image up.

Now I was trying to figure out, what the equivalent for the magick++ solution above is. But neither OverCompositeOp is the right one (only get the overlay image) nor PlusCompositeOp is correct, since it darkens only.

And reading through the Image composition methods in the documentation was not leading to success.

In case you need some sample images, just let me know.

Thank you very much again for all your help!

Olli

[el_supremo]

Code: Select all

composite image.png overlay.png -compose Overlay final.png

The order of the input files in the composite command is "src" then "dest", whereas the convert command and the Magick++ (or MagickWand) functions use "dest" then "src".
In MagickWand the composite function is called like this:

Code: Select all

	// This does the src (overlay) over the dest (background)
	MagickCompositeImage(dest,src,OverlayCompositeOp,0,0);

So try switching the order of the input files. If this doesn't help, post example files and I'll have another look.


Pete

[The_Chaos]

Hi again Pete,

sorry It took so long to answer, but I had quite some things to work on.

I checked again with the order of the composite command, and it doesn't work for me.
so I attached the image files used and also the final files that come out of the command line version (final_cmd.png) and the c++ version (final_cpp.png).

The command line version which works for me as expected is:

Code: Select all

convert poster.png \( source.jpg -virtual-pixel transparent -distort Perspective '0,0 108,0 0,800 108,356 600,800 380,356 600,0 380,0' \) poster_mask.png  -composite - |  composite - poster_overlay5.png -compose Overlay final_cmd.png

and this one results in this image:


You can see the "background image" has lights and shadows at the persons side.


The C++ Code I used is:

Code: Select all

#include <Magick++.h>
#include <iostream>
using namespace std;
using namespace Magick;
int main(int argc,char **argv)
{
        // create image objects for use
        Image src, dest, mask, overlay;

        try {
                // define distorsion points
                double distortParams[] = {0,0, 108,0, 0,800, 108,356, 600,800, 380,356, 600,0, 380,0};

                // read image files into objects
                dest.read( "poster.png" );
                mask.read( "poster_mask.png" );
                overlay.read( "poster_overlay5.png" );

                src.read( "source.jpg" );

                // set surrounding pixels to transparent
                src.virtualPixelMethod(TransparentVirtualPixelMethod);

                // distort the src image to fit inside the masked area
                src.distort(MagickCore::PerspectiveDistortion, 16, distortParams, 0);

                // negate the mask, since we need black pixels where content has to be seen
                // although the mask is made with white pixels where content shines through
                mask.negate();

                // combine the destination image and the mask at the origin
                Geometry placement(0,0,0,0);
                dest.clipMask(mask);
                dest.composite( src, placement, OverCompositeOp);


                // Put the overlay on top of the image. 
                overlay.composite( dest, placement, OverCompositeOp);

                // Write the destination image to a file
                overlay.write( "final_cpp.png" );
        }
        catch( Exception &error_ )
        {
                cout << "Caught exception: " << error_.what() << endl;
                return 1;
        }
        return 0;
}

which produces this image:


which doesn't contain the lights/shadows from the overlay image.

So I would love to have the c++ version to generate the same image with the overlay like final_cmd.png.

Maybe you find out what the problem is or probably it works better with the magicwand version.

You will find the complete file archive here.


I would appreciate any help on this thing



Greetings,
Olli

[el_supremo]

Hi Olli,
Your second composite should use OverlayCompositeOp.

Pete

[The_Chaos]

Hi again Pete,

thank you very much for your fast help... that was really easy!
I somehow didn't find this one, don't know why.


Once again I really appreciate your help!


Greetings,
Olli