16位图像的信噪比计算

graduate

signed int temp;
    double mse = 0.0 ;
    for( y = 0; y < c;  y++ )
    for( x = 0; x < r;  x++ )
    {
        temp = (signed int)map1[y][x]-(signed int)map2[y][x];
        mse += (double)temp*temp;
    }
以上计算好一个16位深度图像的均方误差mse之后

信噪比的计算公式是
mse = mse / (1.0*r*c);
double psnr = 10 * log10(65535.0*65535.0/mse);
还是
mse = mse / (1.0*r*c);
double psnr = 10 * log10(255.0*255.0/mse); 呢 ??