Working with Non-Rectangular Shapes in Photo Mosaic Effects

You are here:

I started by loading the original movie brush as the source movie, and setting the canvas size to be the same size as the source movie. I then erased the canvas to the source.

I then used the Local Growth Option for the Canvas interactive Selection (2nd gallery image above). Note that I moused down near the edge of the canvas in the white background (an area that is white in each individual movie frame, and then pulled the cursor slightly to the right to select the white background and pull the selection a little bit into the interior of the object edges as well (3rd gallery image).

I used local rather than global growth for generating the selection because I did not want to select any light areas in the interior of the objects.

I then inverted the selection (4th gallery image). I then ran 2 different application of selection Contraction (5th gallery image). Both inversion and contraction of the current selection are Canvas : Selection menu commands as shown. Contracting the selection makes it slightly smaller. So inverting the selection made the original manually selected exterior of the objects invert into a selection mask for the object in the frame. And contracting the selection helped get rifd of the white fringe issue at the edges of the objects, since we shrinking the object selection outline to get rid of those pixels.

I then ran the Action : Process with Paint Action Sequence : Source to Movie menu comamnd to process the original movie brush movie file loaded in the source area into a new movie brush with an embedded alpha channel containing object masks.

Before I ran that command, I went to the movie preferences dialog (6th gallery image above), and made sure the Embedded Alpha movie preference was set to the Selection option. Doing so means that the embedded alpha channel for each frame will be whatever is in the current selection buffer when the PASeq renders out a finsihed output movie frame. Don’t forget this set of this procedure is not going to work properly, since we’re building a mask in the selection buffer we want to be the embedded alpha channel for each frame.

I also made sure I was using an output movie coded that supports an embedded alpha channel. I used the animation codec with the Millions  of Colors + Depth option (as shown in the 7th gallery image above). Not all movie codecs support embedded alpha channels.

At this point I had a new movie brush file that I could load into my original movie brush paint synthesizer preset. I could then use the Source Alpha Brush option for the Brush Type, as discussed above to generate my movie brush paint nibs. So the shape and Fill From-Fill To blend of the paint nib is defined by the movie brush’s embedded alpha channel, while the Fill From parameter is set to the Brush Image so that the individual objects in the movie brush draw without any white fringing. And Fill To is set to the Canvas Image so the edges of the embedded alpha channel will blend to the existing canvas at the edges of the object paint nibs.

I hope the detailed explanation above gives you some insight into why white fringes were originally seen with this particular movie brush and the paint synthesizer settings that were used to draw with it. And the approach we took to eliminate the white fringing effects. Also, I hope the explanation gives you a better understanding into how the Source Brush (be it computational or image or movie source brush) interacts with the Paint Fill Setup control panel settings to define a paint nib for painting.


Share with Google+
Go to Top