以前写过两篇关于快速高斯滤波的文章，但是代码都没写完整。
算法来源及介绍请参考博主Imageshop的博文http://www.cnblogs.com/Imageshop/archive/2013/01/07/2849782.html。
为了避免需要的朋友在看我的文章时过于浪费时间，所以删除前两篇文章，重写此篇目的为需要的朋友提供有用的信息。
下面给出的是功能完整的 c代码，下面代码无论是格式还是速度都可以进一步优化，请自行处理
/****************************************
* src : 原始图像数据 *
* dst : 模糊后图像数据 *
* width : 图像宽 *
* height : 图像高 *
* stride : 图像每一行字节数 *
* chan : 图像通道数 *
* sigma : 高斯参数 *
* chan : 图像通道数 *
*****************************************/
void IMG_GaussBlur(unsigned char* src, unsigned char* dst, int width, int height, int stride, int chan, float sigma)
{
int i = 0;
int h,w;
int row = 0;
int col = 0;
int pos = 0;
int channel = 0;
int n = 0;
int bufsize = 0;
int size = 0;
int rowstride = 0;
int itemp0 = 0;
int itemp1 = 0;
float temp = 0;
float fTab[256] = {0};
unsigned char* ps;
float *pIn;
int blurH = height+3;
int blurW = width+3;
for (i=0; i<256; i++)
{
fTab[i] = i+1;
}
int channelsize = width*height+(width+height)*6;
if (width>height)
{
bufsize = width+6;
}
else
{
bufsize = height+6;
}
float* w1 = (float *) malloc (bufsize * sizeof (float));
float *w2 = (float *) malloc (bufsize * sizeof (float));
float *in = (float *) malloc (channelsize * sizeof (float));
float *out = (float *) malloc (channelsize * sizeof (float));
//----------------计算高斯核---------------------------------------//
float q = 0;
float q2, q3;
double b0;
double b1;
double b2;
double b3;
double B = 0;
int N = 3;
if (sigma >= 2.5)
{
q = 0.98711 * sigma - 0.96330;
}
else if ((sigma >= 0.5) && (sigma < 2.5))
{
q = 3.97156 - 4.14554 * (float) sqrt ((double) 1 - 0.26891 * sigma);
}
else
{
q = 0.1147705018520355224609375;
}
q2 = q * q;
q3 = q * q2;
b0 = (1.57825+(2.44413*q)+(1.4281 *q2)+(0.422205*q3));
b1 = ( (2.44413*q)+(2.85619*q2)+(1.26661 *q3));
b2 = ( -((1.4281*q2)+(1.26661 *q3)));
b3 = ( (0.422205*q3));
B = 1.0-((b1+b2+b3)/b0);
//加速方法 减少循环多次/b0
b1 /= b0;
b2 /= b0;
b3 /= b0;
//----------------计算高斯核结束---------------------------------------//
// 处理图像的多个通道
for (channel = 0; channel < chan; channel++)
{
// 获取一个通道的所有像素值
pIn = in;
for (h=0; h<height; h++)
{
ps = src + h*stride;
for (w=0;w<width;w++)
{
/* 0-255 => 1-256 */
*pIn++ = fTab[ps[channel]];
if (w==width-1)
{
*pIn++ = fTab[ps[channel]];
*pIn++ = fTab[ps[channel]];
*pIn++ = fTab[ps[channel]];
}
ps+=chan;
}
}
memcpy(in+(height)*(width+3), in+(height-1)*(width+3), (width+3)*sizeof(float));
memcpy(in+(height+1)*(width+3), in+(height-1)*(width+3), (width+3)*sizeof(float));
memcpy(in+(height+2)*(width+3), in+(height-1)*(width+3), (width+3)*sizeof(float));
//纵向处理
size = blurW;
bufsize = size+3;
size -= 1;
for (row=0 ;row < blurH; row++)
{
pos = row * blurW;
temp = (in + pos)[0];
w1[0] = temp;
w1[1] = temp;
w1[2] = temp;
for ( i = 0 , n=3; i <= size ; i++, n++)
{
w1[n] = (float)(B*(in + pos)[i] + ((b1*w1[n-1] + b2*w1[n-2] + b3*w1[n-3] )));
}
temp = w1[size+3];
w2[size+1]= temp;
w2[size+2]= temp;
w2[size+3]= temp;
for (i = size, n = i; i >= 0; i--, n--)
{
(out + pos)[i] = w2[n] = (float)(B*w1[n] + ((b1*w2[n+1] + b2*w2[n+2] + b3*w2[n+3] )));
}
}
//横向处理
size = blurH;
rowstride = blurW;
bufsize = size+3;
size -= 1;
for (col=0; col < blurW; col++)
{
temp = (out + col)[0];
w1[0] = temp;
w1[1] = temp;
w1[2] = temp;
for ( i = 0 , n=3; i <= size ; i++, n++)
{
w1[n] = (float)(B*(out + col)[i*rowstride] + ((b1*w1[n-1] + b2*w1[n-2] + b3*w1[n-3] )));
}
temp = w1[size+3];
w2[size+1] = temp;
w2[size+2] = temp;
w2[size+3] = temp;
for (i = size, n = i; i >= 0; i--, n--)
{
(in + col)[i * rowstride] =w2[n]= (float)(B*w1[n] + ((b1*w2[n+1] + b2*w2[n+2] + b3*w2[n+3] )));
}
}
//修正偏移的拷贝方法
for(int y=0; y<height; y++)
{
ps = dst+ y*stride;
itemp1 = (y+3)*blurW; // +3
for (int x=0; x<width; x++)
{
ps[channel] = in[itemp1+x+3]-1;
ps+=chan;
}
}
}
free (w1);
free (w2);
free (in);
free (out);
}
调用参考
IplImage* src = cvLoadImage("./test.jpg", 1); IplImage* dst = cvLoadImage("./test.jpg", 1); IMG_GaussBlur((unsigned char*)src->imageData, (unsigned char*)dst->imageData, src->width, src->height, src->widthStep, src->nChannels, 2);
原文：http://www.cnblogs.com/utopiaT/p/4549936.html