isn't it obvious? "pixel average" tells you everything you need to know.

for each pixel in the various images just separate the r, g and b components, add them up and divide by the number. then use this as the r, g, b of the destination image. and use an int for the sum because a byte will overflow.

depends on the source images. that just looks like a bunch of frames from a tracking shot of a taxi. taxi is similar in all shots, the background changes.

jianqiu, there's really not a lot to it, is probably 20 lines of code. look up the following in the reference and examples.

loadImage();
for();      // looping
get();      // gets pixel value
red() / green() / blue() // extract colour values

(pixels[] would be faster, and just repeated calls to blend(..., BLEND); would do it)