This is my first test repo on GitHub:-)
It contains simple, neat (C++ template) functions to transform from yuv(nv21, nv12 is also possible) to rgb(bgr, other formats are easy to extend)

following are comparing results(1280x720 nv21) with or without arm neon instructions:

iPhone3GS : 55.45 vs 13.34(neon) milliseconds
iPhone4S : 30.76 vs 10.43(neon) milliseconds
hTC ruby : 32.25 vs 15.33(neon) milliseconds