对于YUV422的格式，网上有一大堆资料，这里就不说了。直奔主题，给出如何转换的函数，一如既往，只用代码说事。
YUV422有打包格式(Packed)，一如前文所述。同时还有平面格式(Planar)，即Y、U、V是分开存储的，每个分量占一块地方，其中Y为width*height，而U、V合占width*height，该种格式每个像素占16比特。根据U、V的顺序，分出2种格式，U前V后即YUV422P，也叫I422，V前U后，叫YV16(YV表示Y后面跟着V，16表示16bit)。另外，还有一种变态的半平面格式(Semi-planar)，即Y单独占一块地方，但其后U、V又紧挨着排在一起，根据U、V的顺序，又有2种，U前V后叫NV16，在国内好像很多人叫它为YUV422SP格式；V前U后叫NV61。不过这种格式似乎不太受VLC欢迎(具体可去看看VLC的wiki)。
先给出YUV422平面格式的转换函数，如下：

/**
内存分布
                    w
            +--------------------+
            |Y0Y1Y2Y3...         |
            |...                 |   h
            |...                 |
            |                    |
            +--------------------+
            |U0U1      |
            |...       |   h
            |...       |
            |          |
            +----------+
            |V0V1      |
            |...       |  h
            |...       |
            |          |
            +----------+
                w/2
*/
void yuv422p_to_rgb24(YUV_TYPE type, unsigned char* yuv422p, unsigned char* rgb, int width, int height)
{
    int y, cb, cr;
    int r, g, b;
    int i = 0;
    unsigned char* p_y;
    unsigned char* p_u;
    unsigned char* p_v;
    unsigned char* p_rgb;

    p_y = yuv422p;
    p_u = p_y + width * height;
    p_v = p_u + width * height / 2;

    if (type == FMT_YV16)
    {
        p_v = p_y + width * height;
        p_u = p_u + width * height / 2;
    }
    p_rgb = rgb;

    init_yuv422p_table();

    for (i = 0; i < width * height / 2; i++)
    {
        y  = p_y[0];
        cb = p_u[0];
        cr = p_v[0];

        r = MAX (0, MIN (255, (V[cr] + Y1[y])/10000));   //R value
        b = MAX (0, MIN (255, (U[cb] + Y1[y])/10000));   //B value
        g = MAX (0, MIN (255, (Y2[y] - 5094*(r) - 1942*(b))/10000)); //G value

        // 此处可调整RGB排序，BMP图片排序为BGR
        // 默认排序为：RGB
        p_rgb[0] = r;
        p_rgb[1] = g;
        p_rgb[2] = b;

        y  = p_y[1];
        cb = p_u[0];
        cr = p_v[0];
        r = MAX (0, MIN (255, (V[cr] + Y1[y])/10000));   //R value
        b = MAX (0, MIN (255, (U[cb] + Y1[y])/10000));   //B value
        g = MAX (0, MIN (255, (Y2[y] - 5094*(r) - 1942*(b))/10000)); //G value

        p_rgb[3] = r;
        p_rgb[4] = g;
        p_rgb[5] = b;

        p_y += 2;
        p_u += 1;
        p_v += 1;
        p_rgb += 6;
    }
}

接着给出NV16、NV61转换成RGB的函数，如下：

/**
内存分布
                    w
            +--------------------+
            |Y0Y1Y2Y3...         |
            |...                 |   h
            |...                 |
            |                    |
            +--------------------+
            |U0V0U1V1            |
            |...                 |   h
            |...                 |
            |                    |
            +--------------------+
                w/2
UV交织为NV16，VU交织为NV61
可以与上一函数合并，但方便查看，还是不合并
*/
void yuv422sp_to_rgb24(YUV_TYPE type, unsigned char* yuv422sp, unsigned char* rgb, int width, int height)
{
    int y, cb, cr;
    int r, g, b;
    int i = 0;
    unsigned char* p_y;
    unsigned char* p_uv;
    unsigned char* p_rgb;

    p_y = yuv422sp;
    p_uv = p_y + width * height;    // uv分量在Y后面
    p_rgb = rgb;

    init_yuv422p_table();

    for (i = 0; i < width * height / 2; i++)
    {
        y  = p_y[0];
        if (type ==  FMT_NV16)
        {
            cb = p_uv[0];
            cr = p_uv[1];    // v紧跟u，在u的下一个位置
        }
        if (type == FMT_NV61)
        {
            cr = p_uv[0];
            cb = p_uv[1];    // u紧跟v，在v的下一个位置
        }

        r = MAX (0, MIN (255, (V[cr] + Y1[y])/10000));   //R value
        b = MAX (0, MIN (255, (U[cb] + Y1[y])/10000));   //B value
        g = MAX (0, MIN (255, (Y2[y] - 5094*(r) - 1942*(b))/10000)); //G value

        // 此处可调整RGB排序，BMP图片排序为BGR
        // 默认排序为：RGB
        p_rgb[0] = r;
        p_rgb[1] = g;
        p_rgb[2] = b;

        y  = p_y[1];
        if (type ==  FMT_NV16)
        {
            cb = p_uv[0];
            cr = p_uv[1];
        }
        if (type ==  FMT_NV61)
        {
            cr = p_uv[0];
            cb = p_uv[1];
        }

        r = MAX (0, MIN (255, (V[cr] + Y1[y])/10000));   //R value
        b = MAX (0, MIN (255, (U[cb] + Y1[y])/10000));   //B value
        g = MAX (0, MIN (255, (Y2[y] - 5094*(r) - 1942*(b))/10000)); //G value

        p_rgb[3] = r;
        p_rgb[4] = g;
        p_rgb[5] = b;

        p_y += 2;
        p_uv += 2;
        p_rgb += 6;
    }
}

参考资料：

李迟 2015.8.5 晚上

近来经济拮据，如本文对阁下有帮助，可慷慨解囊赞助笔者以输出更多好文章。
[email protected] 或 微信fly_camel_fly 均可。感谢！
                                

本文固定链接: /my-study/yuv-learning-yuv422p-to-rgb24.html