SSE算法优化系列4：使用SSE优化Sobel边缘检测算法，加速比9.5倍

前言

仍然是学习Imageshop的文章，https://www.cnblogs.com/Imageshop/p/7285564.html 。做了一个小总结，并且完整实现了这个SSE优化的算法，可以关注我专门用SIMD优化图像处理算法的工程：https://github.com/BBuf/Image-processing-algorithm-Speed

传统的Sobel算法实现

我之前写过Sobel边缘检测算法如何产生X方向和方向的系数，https://blog.youkuaiyun.com/just_sort/article/details/85015054 。具体原理就不再赘述，这里原理就不再叙述了，给出指针版本实现的边缘检测算法。

inline unsigned char IM_ClampToByte(int Value)
{
	if (Value < 0)
		return 0;
	else if (Value > 255)
		return 255;
	else
		return (unsigned char)Value;
	//return ((Value | ((signed int)(255 - Value) >> 31)) & ~((signed int)Value >> 31));
}

void Sobel_FLOAT(unsigned char *Src, unsigned char *Dest, int Width, int Height, int Stride) {
	int Channel = Stride / Width;
	unsigned char *RowCopy = (unsigned char*)malloc((Width + 2) * 3 * Channel);
	unsigned char *First = RowCopy;
	unsigned char *Second = RowCopy + (Width + 2) * Channel;
	unsigned char *Third = RowCopy + (Width + 2) * 2 * Channel;
	//拷贝第二行数据，边界值填充
	memcpy(Second, Src, Channel);
	memcpy(Second + Channel, Src, Width*Channel);
	memcpy(Second + (Width + 1)*Channel, Src + (Width - 1)*Channel, Channel);
	//第一行和第二行一样
	memcpy(First, Second, (Width + 2) * Channel);
	//拷贝第三行数据，边界值填充
	memcpy(Third, Src + Stride, Channel);
	memcpy(Third + Channel, Src + Stride, Width * Channel);
	memcpy(Third + (Width + 1) * Channel, Src + Stride + (Width - 1) * Channel, Channel);

	for (int Y = 0; Y < Height; Y++) {
		unsigned char *LinePS = Src + Y * Stride;
		unsigned char *LinePD = Dest + Y * Stride;
		if (Y != 0) {
			unsigned char *Temp = First;
			First = Second;