概述
项目中用到yuv NV12转BGR24的算法，总结了几种常用算法代码如下。
直接转换
//NV 12 转BGR
void NV12_T_BGR(unsigned int width, unsigned int height, unsigned char *yuyv,
unsigned char *bgr) {
const int nv_start = width * height;
int i, j, index = 0, rgb_index = 0;
unsigned char y, u, v;
int r, g, b, nv_index = 0;
for (i = 0; i < height; i++) {
for (j = 0; j < width; j++) {
//nv_index = (rgb_index / 2 - width / 2 * ((i + 1) / 2)) * 2;
nv_index = i / 2 * width + j - j % 2;
y = yuyv[rgb_index];
v = yuyv[nv_start + nv_index];
u = yuyv[nv_start + nv_index + 1];
// u = yuyv[nv_start + nv_index ];
// v = yuyv[nv_start + nv_index + 1];
r = y + (140 * (v - 128)) / 100; //r
g = y - (34 * (u - 128)) / 100 - (71 * (v - 128)) / 100; //g
b = y + (177 * (u - 128)) / 100; //b
if (r > 255)
r = 255;
if (g > 255)
g = 255;
if (b > 255)
b = 255;
if (r < 0)
r = 0;
if (g < 0)
g = 0;
if (b < 0)
b = 0;
index = rgb_index % width + (height - i - 1) * width;
bgr[index * 3 + 2] = r;
bgr[index * 3 + 1] = g;
bgr[index * 3 + 0] = b;
rgb_index++;
}
}
}直接转换（查表优化）
static int crv_tab[256];
static int cbu_tab[256];
static int cgu_tab[256];
static int cgv_tab[256];
static int tab_76309[256];
static unsigned char clp[1024]; //for clip in CCIR601
static int inityuv420p = 0;
void init_yuv420p_table()
{
int crv,cbu,cgu,cgv;
int i,ind;
if (inityuv420p == 1) return;
crv = 104597; cbu = 132201; /* fra matrise i global.h */
cgu = 25675; cgv = 53279;
for (i = 0; i < 256; i++)
{
crv_tab[i] = ((i-128) * crv);
cbu_tab[i] = ((i-128) * cbu);
cgu_tab[i] = ((i-128) * cgu);
cgv_tab[i] = ((i-128) * cgv);
tab_76309[i] = ((76309*(i-16)));
}
for (i = 0; i < 384; i++)
clp[i] = 0;
ind = 384;
for (i = 0;i < 256; i++)
clp[ind++] = i;
ind = 640;
for (i = 0;i < 384; i++)
clp[ind++] = 255;
inityuv420p = 1;
}
void yuv420sp_to_rgb24(int type, unsigned char* yuvbuffer,unsigned char* rgbbuffer, int width,int height)
{
if (inityuv420p != 1) init_yuv420p_table();
int y1, y2, u = 0, v = 0;
unsigned char *py1, *py2;
int i, j, c1, c2, c3, c4;
unsigned char *d1, *d2;
unsigned char *src_u;
src_u = yuvbuffer + width * height; // u
py1 = yuvbuffer; // y
py2 = py1 + width;
d2 = rgbbuffer + 3 * width * (height - 2);
d1 = d2 + 3 * width;
int dotline = 9 * width;
static unsigned char *pclp384 = clp + 384;
for (j = 0; j < height; j += 2)
{
for (i = 0; i < width; i += 2)
{
u = *src_u++;
v = *src_u++; // v紧跟u，在u的下一个位置
c1 = crv_tab[v];
//c2 = cgu_tab[u];
//c3 = cgv_tab[v];
c4 = cbu_tab[u];
int c2c3 = cgu_tab[u] + cgv_tab[v];
//RGB
//up-left
/*y1 = tab_76309[*py1++];
*d1++ = clp[384+((y1 + c1)>>16)];
*d1++ = clp[384+((y1 - c2 - c3)>>16)];
*d1++ = clp[384+((y1 + c4)>>16)];
//down-left
y2 = tab_76309[*py2++];
*d2++ = clp[384+((y2 + c1)>>16)];
*d2++ = clp[384+((y2 - c2 - c3)>>16)];
*d2++ = clp[384+((y2 + c4)>>16)];
//up-right
y1 = tab_76309[*py1++];
*d1++ = clp[384+((y1 + c1)>>16)];
*d1++ = clp[384+((y1 - c2 - c3)>>16)];
*d1++ = clp[384+((y1 + c4)>>16)];
//down-right
y2 = tab_76309[*py2++];
*d2++ = clp[384+((y2 + c1)>>16)];
*d2++ = clp[384+((y2 - c2 - c3)>>16)];
*d2++ = clp[384+((y2 + c4)>>16)];*/
y1 = tab_76309[*py1++];
*d1++ = pclp384[((y1 + c1))>>16];
*d1++ = pclp384[((y1 - c2c3))>>16];
*d1++ = pclp384[((y1 + c4))>>16];
//down-left
y2 = tab_76309[*py2++];
*d2++ = pclp384[((y2 + c1))>>16];
*d2++ = pclp384[((y2 - c2c3))>>16];
*d2++ = pclp384[((y2 + c4))>>16];
//up-right
y1 = tab_76309[*py1++];
*d1++ = pclp384[((y1 + c1))>>16];
*d1++ = pclp384[((y1 - c2c3))>>16];
*d1++ = pclp384[((y1 + c4))>>16];
//down-right
y2 = tab_76309[*py2++];
*d2++ = pclp384[((y2 + c1))>>16];
*d2++ = pclp384[((y2 - c2c3))>>16];
*d2++ = pclp384[((y2 + c4))>>16];
}
d1 -= dotline;
d2 -= dotline;
py1 += width;
py2 += width;
}
}Neon优化
#include "arm_neon.h"
const uint8_t Y_SUBS[8] = { 16, 16, 16, 16, 16, 16, 16, 16 };
const uint8_t UV_SUBS[8] = { 128, 128, 128, 128, 128, 128, 128, 128 };
void yv12_to_rgb24_neon(unsigned char *src, int width, int height, unsigned char *RGBOut) //仅仅支持宽度为16的倍数的数据帧
{
int i, j;
int nWH = width * height;
unsigned char *pY1 = src;
unsigned char *pY2 = src + width;
unsigned char *pUV = src + nWH;
uint8x8_t Y_SUBvec = vld1_u8(Y_SUBS);
uint8x8_t UV_SUBvec = vld1_u8(UV_SUBS);
//int width2 = width >> 1;
int width3 = (width << 2) - width;
int width9 = (width << 3) + width;
//unsigned char *RGBOut1 = RGBOut;
// unsigned char *RGBOut2 = RGBOut1 + width3;
unsigned char *RGBOut1 = RGBOut + 3 * width * (height - 2);
unsigned char *RGBOut2 = RGBOut1 + width3;
unsigned char tempUV[8];
// YUV 4:2:0
for (j = 0; j < height; j += 2)
{
for (i = 0; i < width; i += 8)
{
tempUV[0] = pUV[0];
tempUV[1] = pUV[2];
tempUV[2] = pUV[4];
tempUV[3] = pUV[6];
tempUV[4] = pUV[1];
tempUV[5] = pUV[3];
tempUV[6] = pUV[5];
tempUV[7] = pUV[7];
pUV += 8;
uint8x8_t nUVvec = vld1_u8(tempUV);
int16x8_t nUVvec16 = vmovl_s8(vsub_s8(nUVvec, UV_SUBvec));//减后区间-128到127
int16x4_t V_4 = vget_low_s16(nUVvec16);
int16x4x2_t V16x4x2 = vzip_s16(V_4, V_4);
//int16x8_t V16x8_;
//memcpy(&V16x8_, &V16x4x2, 16);
//int16x8_t* V16x8 = (int16x8_t*)(&V16x8_);
int16x8_t* V16x8 = (int16x8_t*)(&V16x4x2);
int16x4_t U_4 = vget_high_s16(nUVvec16);
int16x4x2_t U16x4x2 = vzip_s16(U_4, U_4);
int16x8_t* U16x8 = (int16x8_t*)(&U16x4x2);
//公式1
int16x8_t VV1 = vmulq_n_s16(*V16x8, 102);
int16x8_t UU1 = vmulq_n_s16(*U16x8, 129);
int16x8_t VVUU1 = vmlaq_n_s16(vmulq_n_s16(*V16x8, 52), *U16x8, 25);
uint8x8_t nYvec;
uint8x8x3_t RGB;
uint16x8_t Y16;
//上行
nYvec = vld1_u8(pY1);
pY1 += 8;
//公式1
Y16 = vmulq_n_u16(vmovl_u8(vqsub_u8(nYvec, Y_SUBvec)), 74);//公式1
RGB.val[0] = vqmovun_s16(vshrq_n_s16(vaddq_u16(Y16, UU1), 6));
RGB.val[1] = vqmovun_s16(vshrq_n_s16(vsubq_u16(Y16, VVUU1), 6));
RGB.val[2] = vqmovun_s16(vshrq_n_s16(vaddq_u16(Y16, VV1), 6));
vst3_u8(RGBOut1, RGB);
RGBOut1 += 24;
//下行
nYvec = vld1_u8(pY2);
pY2 += 8;
//公式1
Y16 = vmulq_n_u16(vmovl_u8(vqsub_u8(nYvec, Y_SUBvec)), 74);//公式1
RGB.val[0] = vqmovun_s16(vshrq_n_s16(vaddq_u16(Y16, UU1), 6));
RGB.val[1] = vqmovun_s16(vshrq_n_s16(vsubq_u16(Y16, VVUU1), 6));
RGB.val[2] = vqmovun_s16(vshrq_n_s16(vaddq_u16(Y16, VV1), 6));
vst3_u8(RGBOut2, RGB);
RGBOut2 += 24;
}
//RGBOut1 += width3;
//RGBOut2 += width3;
RGBOut1 -= width9;
RGBOut2 -= width9;
pY1 += width;
pY2 += width;
}
}使用ffmpeg中的转换函数实现
//ffmpeg变量
AVPicture pFrameYUV,pFrameBGR;
uint8_t * ptmp;
struct SwsContext* imgCtx = NULL;
void Init(){
//初始化ffmpeg
{
//LOGI_MATRIX(“Start->Sws”);
imgCtx = sws_getContext(instance->pWidth,instance->pHeight,AV_PIX_FMT_NV21,instance->pWidth,instance->pHeight,PIX_FMT_BGR24,SWS_BILINEAR,0,0,0);
}
}
void YV12ToBGR24_FFmpeg(unsigned char* pYUV,unsigned char* pBGR24,int width,int height)
{
if (width < 1 || height < 1 || pYUV == NULL || pBGR24 == NULL)
return ;
//int srcNumBytes,dstNumBytes;
//uint8_t *pSrc,*pDst;
//LOGI_MATRIX("Start->trans");
//pFrameYUV = avpicture_alloc();
//srcNumBytes = avpicture_get_size(PIX_FMT_YUV420P,width,height);
//pSrc = (uint8_t *)malloc(sizeof(uint8_t) * srcNumBytes);
// LOGI_MATRIX("Start->avpicture_fill");
avpicture_fill(&pFrameYUV,pYUV,AV_PIX_FMT_NV12,instance->pWidth,instance->pHeight);
//U,V互换
// ptmp =pFrameYUV.data[1];
// pFrameYUV.data[1]=pFrameYUV.data[2];
// pFrameYUV.data [2]=ptmp;
//pFrameBGR = avcodec_alloc_frame();
//dstNumBytes = avpicture_get_size(PIX_FMT_BGR24,width,height);
//pDst = (uint8_t *)malloc(sizeof(uint8_t) * dstNumBytes);
// LOGI_MATRIX("Start->avpicture_fill");
if (imgCtx != NULL){
// LOGI_MATRIX("imgCtx->Sws");
sws_scale(imgCtx,pFrameYUV.data,pFrameYUV.linesize,0,height,pFrameBGR.data,pFrameBGR.linesize);
//
//if(imgCtx){
// sws_freeContext(imgCtx);
// imgCtx = NULL;
// }
return ;
}
else{
LOGI_MATRIX("imgCtx==NULL");
//sws_freeContext(imgCtx);
// imgCtx = NULL;
return ;
}
}ffmpeg编译测试博客链接
希望对您有所帮助！
