I have found using the PVector [] works great!

Example

color c = color(255,0,0);
int w = 255;
int h = 255;

int all = w*h;

PVector [] scr = new PVector [all];

void setup(){
size(255,255);

for (int i =0;i<all;i++){

scr[i]=new PVector (255,0,0);
set(int(i/w),int(i-(i/w),color(scr[i].x.scr[i].y,scr[i].z));
}

The possibilities are endless. I like this method so I don't have to use red(),green() or blue() or even bit shifting (which is faster btw)

Why would you want to save integer information in float information? Do you want one float per pixel or per color component?