__global__ void rgb_to_ycrcb(uchar3 *input,uchar3 *out,int h,int w,float *y)
{
unsigned int x=blockDim.x*blockIdx.x+threadIdx.x;
unsigned int y=blockDim.y*blockIdx.y+threadIdx.y;
if(x<w&&y<h)
{
uchar3 rgb=input[y*w+x];
yy[y*w+x]=(16 + 0.257 * rgb.x + 0.504 * rgb.y + 0.098 * rgb.z)/255.0;
if(yy[y*w+x]<0) yy