how to reduce opacity of mask(transparent image) borders in Processing?

Question

I am going to show 2 layers(depth image from kinect and a mask,respectively) in my processing project. In this project, I am going to hide areas which there is no user available by utilizing a mask. Note that, "mask" gets its pixels from a movie named "movie2" in my code.
Is there any way to blur borders of users by reducing opacity of mask borders?

Here is my code:

    import SimpleOpenNI.*;
PImage mask,depth;
SimpleOpenNI kinect;
int[] userMap;

void setup(){
  kinect = new SimpleOpenNI(this);
  size(640,480);
  kinect.enableDepth();
  kinect.enableUser();
      }

      void draw(){
        kinect.update();
        depth=kinect.depthImage();
        userMap = kinect.userMap();
        image(depth,0,0);
        mask=loadImage("mask.jpg");
        mask.loadPixels();
    for (int y=0; y<480; y++) {
          for (int x=0; x<640; x++) {
            int index = x + y *640;
            if (userMap[index]!= 0) {
             mask.pixels[index]=color(0,0,0,0);
            }
           }
        }
      mask.updatePixels();
      image(mask,0,0);
    }

Answer 1

There are a few things that could be improved:

1. No need to load the mask multiple times a second in draw() (once in setup will be enough)
2. You can make use the mask() function to apply a mask
3. You can manage a custom user image with transparent pixels for masking and opaque pixels for user pixels which you can filter to reduce border (e.g. ERODE or BLUR via filter()

Other notes:

• Processing's built-in blur can be a little slow, but you can use Greg Borensteinn's OpenCV for Processing library which has faster filters (blur,erode)
• If you're only checking the userMap and you don't need x,y coordinates, you can simply use a flat loop instead of a nested loop and gain a bit of speed.

Here's a stab at illustrating the points above as code, but bare in mind this incomplete/untested:

import processing.video.*;

class effect2{
  PImage mask;

  int[] userMap;
  PImage userImage;

  int numPixels;

  effect2(){
   //movie2  = new Movie(this,"MOVIE_2_PATH_HERE");//might be initialised somewhere
   movie2.play();
   //initialize image once 
   userImage = createImage(640,480,ARGB);
   userImage.loadPixels();

   numPixels = userImage.width * userImage.height;

   userMap = new int[numPixels];
  }

  void drawing(){
    //kinect.update(); is called at some point

    image(depth,0,0);

    //update user map
    kinect.userMap(userMap); 
    //update user image based on user map
    //flat (1D) for loop is faster than nested (2D) for loop
    for(int i = 0; i < numPixels; i++){
      if (userMap[i] == 0) {
          userImage.pixels[i] = 0x00000000;//transparent black
        }else{
          userImage.pixels[i] = 0xFFFFFFFF;//opaque white
        }
    }
    userImage.updatePixels();

    //filter/blur as needed
    //I recommend using https://github.com/atduskgreg/opencv-processing for faster blur (and other filters) 
    userImage.filter(BLUR,2);

    //apply mask
    movie2.mask(userImage,0,0);

    //preview mask
    image(userImage,0,0,64,48);
  }

}
//the movie is updated too
void movieEvent(Movie m){
  m.read();
}