几种不同大小的dct变换

于 2021-06-19 11:39:34 发布
阅读量499 收藏 1
点赞数
分类专栏： 压缩编码 文章标签： c++ 音视频
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/qq_34162500/article/details/118051220
版权 
const int16_t g_t4[4][4] =
{
    { 64, 64, 64, 64 },
    { 83, 36, -36, -83 },
    { 64, -64, -64, 64 },
    { 36, -83, 83, -36 }
};

const int16_t g_t8[8][8] =
{
    { 64, 64, 64, 64, 64, 64, 64, 64 },
    { 89, 75, 50, 18, -18, -50, -75, -89 },
    { 83, 36, -36, -83, -83, -36, 36, 83 },
    { 75, -18, -89, -50, 50, 89, 18, -75 },
    { 64, -64, -64, 64, 64, -64, -64, 64 },
    { 50, -89, 18, 75, -75, -18, 89, -50 },
    { 36, -83, 83, -36, -36, 83, -83, 36 },
    { 18, -50, 75, -89, 89, -75, 50, -18 }
};

const int16_t g_t16[16][16] =
{
    { 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 },
    { 90, 87, 80, 70, 57, 43, 25,  9, -9, -25, -43, -57, -70, -80, -87, -90 },
    { 89, 75, 50, 18, -18, -50, -75, -89, -89, -75, -50, -18, 18, 50, 75, 89 },
    { 87, 57,  9, -43, -80, -90, -70, -25, 25, 70, 90, 80, 43, -9, -57, -87 },
    { 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83 },
    { 80,  9, -70, -87, -25, 57, 90, 43, -43, -90, -57, 25, 87, 70, -9, -80 },
    { 75, -18, -89, -50, 50, 89, 18, -75, -75, 18, 89, 50, -50, -89, -18, 75 },
    { 70, -43, -87,  9, 90, 25, -80, -57, 57, 80, -25, -90, -9, 87, 43, -70 },
    { 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64 },
    { 57, -80, -25, 90, -9, -87, 43, 70, -70, -43, 87,  9, -90, 25, 80, -57 },
    { 50, -89, 18, 75, -75, -18, 89, -50, -50, 89, -18, -75, 75, 18, -89, 50 },
    { 43, -90, 57, 25, -87, 70,  9, -80, 80, -9, -70, 87, -25, -57, 90, -43 },
    { 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36 },
    { 25, -70, 90, -80, 43,  9, -57, 87, -87, 57, -9, -43, 80, -90, 70, -25 },
    { 18, -50, 75, -89, 89, -75, 50, -18, -18, 50, -75, 89, -89, 75, -50, 18 },
    {  9, -25, 43, -57, 70, -80, 87, -90, 90, -87, 80, -70, 57, -43, 25, -9 }
};

const int16_t g_t32[32][32] =
{
    { 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 },
    { 90, 90, 88, 85, 82, 78, 73, 67, 61, 54, 46, 38, 31, 22, 13,  4, -4, -13, -22, -31, -38, -46, -54, -61, -67, -73, -78, -82, -85, -88, -90, -90 },
    { 90, 87, 80, 70, 57, 43, 25,  9, -9, -25, -43, -57, -70, -80, -87, -90, -90, -87, -80, -70, -57, -43, -25, -9,  9, 25, 43, 57, 70, 80, 87, 90 },
    { 90, 82, 67, 46, 22, -4, -31, -54, -73, -85, -90, -88, -78, -61, -38, -13, 13, 38, 61, 78, 88, 90, 85, 73, 54, 31,  4, -22, -46, -67, -82, -90 },
    { 89, 75, 50, 18, -18, -50, -75, -89, -89, -75, -50, -18, 18, 50, 75, 89, 89, 75, 50, 18, -18, -50, -75, -89, -89, -75, -50, -18, 18, 50, 75, 89 },
    { 88, 67, 31, -13, -54, -82, -90, -78, -46, -4, 38, 73, 90, 85, 61, 22, -22, -61, -85, -90, -73, -38,  4, 46, 78, 90, 82, 54, 13, -31, -67, -88 },
    { 87, 57,  9, -43, -80, -90, -70, -25, 25, 70, 90, 80, 43, -9, -57, -87, -87, -57, -9, 43, 80, 90, 70, 25, -25, -70, -90, -80, -43,  9, 57, 87 },
    { 85, 46, -13, -67, -90, -73, -22, 38, 82, 88, 54, -4, -61, -90, -78, -31, 31, 78, 90, 61,  4, -54, -88, -82, -38, 22, 73, 90, 67, 13, -46, -85 },
    { 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83, 83, 36, -36, -83, -83, -36, 36, 83 },
    { 82, 22, -54, -90, -61, 13, 78, 85, 31, -46, -90, -67,  4, 73, 88, 38, -38, -88, -73, -4, 67, 90, 46, -31, -85, -78, -13, 61, 90, 54, -22, -82 },
    { 80,  9, -70, -87, -25, 57, 90, 43, -43, -90, -57, 25, 87, 70, -9, -80, -80, -9, 70, 87, 25, -57, -90, -43, 43, 90, 57, -25, -87, -70,  9, 80 },
    { 78, -4, -82, -73, 13, 85, 67, -22, -88, -61, 31, 90, 54, -38, -90, -46, 46, 90, 38, -54, -90, -31, 61, 88, 22, -67, -85, -13, 73, 82,  4, -78 },
    { 75, -18, -89, -50, 50, 89, 18, -75, -75, 18, 89, 50, -50, -89, -18, 75, 75, -18, -89, -50, 50, 89, 18, -75, -75, 18, 89, 50, -50, -89, -18, 75 },
    { 73, -31, -90, -22, 78, 67, -38, -90, -13, 82, 61, -46, -88, -4, 85, 54, -54, -85,  4, 88, 46, -61, -82, 13, 90, 38, -67, -78, 22, 90, 31, -73 },
    { 70, -43, -87,  9, 90, 25, -80, -57, 57, 80, -25, -90, -9, 87, 43, -70, -70, 43, 87, -9, -90, -25, 80, 57, -57, -80, 25, 90,  9, -87, -43, 70 },
    { 67, -54, -78, 38, 85, -22, -90,  4, 90, 13, -88, -31, 82, 46, -73, -61, 61, 73, -46, -82, 31, 88, -13, -90, -4, 90, 22, -85, -38, 78, 54, -67 },
    { 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64, 64, -64, -64, 64 },
    { 61, -73, -46, 82, 31, -88, -13, 90, -4, -90, 22, 85, -38, -78, 54, 67, -67, -54, 78, 38, -85, -22, 90,  4, -90, 13, 88, -31, -82, 46, 73, -61 },
    { 57, -80, -25, 90, -9, -87, 43, 70, -70, -43, 87,  9, -90, 25, 80, -57, -57, 80, 25, -90,  9, 87, -43, -70, 70, 43, -87, -9, 90, -25, -80, 57 },
    { 54, -85, -4, 88, -46, -61, 82, 13, -90, 38, 67, -78, -22, 90, -31, -73, 73, 31, -90, 22, 78, -67, -38, 90, -13, -82, 61, 46, -88,  4, 85, -54 },
    { 50, -89, 18, 75, -75, -18, 89, -50, -50, 89, -18, -75, 75, 18, -89, 50, 50, -89, 18, 75, -75, -18, 89, -50, -50, 89, -18, -75, 75, 18, -89, 50 },
    { 46, -90, 38, 54, -90, 31, 61, -88, 22, 67, -85, 13, 73, -82,  4, 78, -78, -4, 82, -73, -13, 85, -67, -22, 88, -61, -31, 90, -54, -38, 90, -46 },
    { 43, -90, 57, 25, -87, 70,  9, -80, 80, -9, -70, 87, -25, -57, 90, -43, -43, 90, -57, -25, 87, -70, -9, 80, -80,  9, 70, -87, 25, 57, -90, 43 },
    { 38, -88, 73, -4, -67, 90, -46, -31, 85, -78, 13, 61, -90, 54, 22, -82, 82, -22, -54, 90, -61, -13, 78, -85, 31, 46, -90, 67,  4, -73, 88, -38 },
    { 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36, 36, -83, 83, -36, -36, 83, -83, 36 },
    { 31, -78, 90, -61,  4, 54, -88, 82, -38, -22, 73, -90, 67, -13, -46, 85, -85, 46, 13, -67, 90, -73, 22, 38, -82, 88, -54, -4, 61, -90, 78, -31 },
    { 25, -70, 90, -80, 43,  9, -57, 87, -87, 57, -9, -43, 80, -90, 70, -25, -25, 70, -90, 80, -43, -9, 57, -87, 87, -57,  9, 43, -80, 90, -70, 25 },
    { 22, -61, 85, -90, 73, -38, -4, 46, -78, 90, -82, 54, -13, -31, 67, -88, 88, -67, 31, 13, -54, 82, -90, 78, -46,  4, 38, -73, 90, -85, 61, -22 },
    { 18, -50, 75, -89, 89, -75, 50, -18, -18, 50, -75, 89, -89, 75, -50, 18, 18, -50, 75, -89, 89, -75, 50, -18, -18, 50, -75, 89, -89, 75, -50, 18 },
    { 13, -38, 61, -78, 88, -90, 85, -73, 54, -31,  4, 22, -46, 67, -82, 90, -90, 82, -67, 46, -22, -4, 31, -54, 73, -85, 90, -88, 78, -61, 38, -13 },
    {  9, -25, 43, -57, 70, -80, 87, -90, 90, -87, 80, -70, 57, -43, 25, -9, -9, 25, -43, 57, -70, 80, -87, 90, -90, 87, -80, 70, -57, 43, -25,  9 },
    {  4, -13, 22, -31, 38, -46, 54, -61, 67, -73, 78, -82, 85, -88, 90, -90, 90, -90, 88, -85, 82, -78, 73, -67, 61, -54, 46, -38, 31, -22, 13, -4 }
};

static void partialButterfly4(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j;
    int E[2], O[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* E and O */
        E[0] = src[0] + src[3];
        O[0] = src[0] - src[3];
        E[1] = src[1] + src[2];
        O[1] = src[1] - src[2];

        dst[0] = (int16_t)((g_t4[0][0] * E[0] + g_t4[0][1] * E[1] + add) >> shift);
        dst[2 * line] = (int16_t)((g_t4[2][0] * E[0] + g_t4[2][1] * E[1] + add) >> shift);
        dst[line] = (int16_t)((g_t4[1][0] * O[0] + g_t4[1][1] * O[1] + add) >> shift);
        dst[3 * line] = (int16_t)((g_t4[3][0] * O[0] + g_t4[3][1] * O[1] + add) >> shift);

        src += 4;
        dst++;
    }
}

static void partialButterfly8(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[4], O[4];
    int EE[2], EO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* E and O*/
        for (k = 0; k < 4; k++)
        {
            E[k] = src[k] + src[7 - k];
            O[k] = src[k] - src[7 - k];
        }

        /* EE and EO */
        EE[0] = E[0] + E[3];
        EO[0] = E[0] - E[3];
        EE[1] = E[1] + E[2];
        EO[1] = E[1] - E[2];

        dst[0] = (int16_t)((g_t8[0][0] * EE[0] + g_t8[0][1] * EE[1] + add) >> shift);
        dst[4 * line] = (int16_t)((g_t8[4][0] * EE[0] + g_t8[4][1] * EE[1] + add) >> shift);
        dst[2 * line] = (int16_t)((g_t8[2][0] * EO[0] + g_t8[2][1] * EO[1] + add) >> shift);
        dst[6 * line] = (int16_t)((g_t8[6][0] * EO[0] + g_t8[6][1] * EO[1] + add) >> shift);

        dst[line] = (int16_t)((g_t8[1][0] * O[0] + g_t8[1][1] * O[1] + g_t8[1][2] * O[2] + g_t8[1][3] * O[3] + add) >> shift);
        dst[3 * line] = (int16_t)((g_t8[3][0] * O[0] + g_t8[3][1] * O[1] + g_t8[3][2] * O[2] + g_t8[3][3] * O[3] + add) >> shift);
        dst[5 * line] = (int16_t)((g_t8[5][0] * O[0] + g_t8[5][1] * O[1] + g_t8[5][2] * O[2] + g_t8[5][3] * O[3] + add) >> shift);
        dst[7 * line] = (int16_t)((g_t8[7][0] * O[0] + g_t8[7][1] * O[1] + g_t8[7][2] * O[2] + g_t8[7][3] * O[3] + add) >> shift);

        src += 8;
        dst++;
    }
}

static void partialButterfly16(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[8], O[8];
    int EE[4], EO[4];
    int EEE[2], EEO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* E and O */
        for (k = 0; k < 8; k++)
        {
            E[k] = src[k] + src[15 - k];
            O[k] = src[k] - src[15 - k];
        }

        /* EE and EO */
        for (k = 0; k < 4; k++)
        {
            EE[k] = E[k] + E[7 - k];
            EO[k] = E[k] - E[7 - k];
        }

        /* EEE and EEO */
        EEE[0] = EE[0] + EE[3];
        EEO[0] = EE[0] - EE[3];
        EEE[1] = EE[1] + EE[2];
        EEO[1] = EE[1] - EE[2];

        dst[0] = (int16_t)((g_t16[0][0] * EEE[0] + g_t16[0][1] * EEE[1] + add) >> shift);
        dst[8 * line] = (int16_t)((g_t16[8][0] * EEE[0] + g_t16[8][1] * EEE[1] + add) >> shift);
        dst[4 * line] = (int16_t)((g_t16[4][0] * EEO[0] + g_t16[4][1] * EEO[1] + add) >> shift);
        dst[12 * line] = (int16_t)((g_t16[12][0] * EEO[0] + g_t16[12][1] * EEO[1] + add) >> shift);

        for (k = 2; k < 16; k += 4)
        {
            dst[k * line] = (int16_t)((g_t16[k][0] * EO[0] + g_t16[k][1] * EO[1] + g_t16[k][2] * EO[2] +
                                       g_t16[k][3] * EO[3] + add) >> shift);
        }

        for (k = 1; k < 16; k += 2)
        {
            dst[k * line] =  (int16_t)((g_t16[k][0] * O[0] + g_t16[k][1] * O[1] + g_t16[k][2] * O[2] + g_t16[k][3] * O[3] +
                                        g_t16[k][4] * O[4] + g_t16[k][5] * O[5] + g_t16[k][6] * O[6] + g_t16[k][7] * O[7] +
                                        add) >> shift);
        }

        src += 16;
        dst++;
    }
}

static void partialButterfly32(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[16], O[16];
    int EE[8], EO[8];
    int EEE[4], EEO[4];
    int EEEE[2], EEEO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* E and O*/
        for (k = 0; k < 16; k++)
        {
            E[k] = src[k] + src[31 - k];
            O[k] = src[k] - src[31 - k];
        }

        /* EE and EO */
        for (k = 0; k < 8; k++)
        {
            EE[k] = E[k] + E[15 - k];
            EO[k] = E[k] - E[15 - k];
        }

        /* EEE and EEO */
        for (k = 0; k < 4; k++)
        {
            EEE[k] = EE[k] + EE[7 - k];
            EEO[k] = EE[k] - EE[7 - k];
        }

        /* EEEE and EEEO */
        EEEE[0] = EEE[0] + EEE[3];
        EEEO[0] = EEE[0] - EEE[3];
        EEEE[1] = EEE[1] + EEE[2];
        EEEO[1] = EEE[1] - EEE[2];

        dst[0] = (int16_t)((g_t32[0][0] * EEEE[0] + g_t32[0][1] * EEEE[1] + add) >> shift);
        dst[16 * line] = (int16_t)((g_t32[16][0] * EEEE[0] + g_t32[16][1] * EEEE[1] + add) >> shift);
        dst[8 * line] = (int16_t)((g_t32[8][0] * EEEO[0] + g_t32[8][1] * EEEO[1] + add) >> shift);
        dst[24 * line] = (int16_t)((g_t32[24][0] * EEEO[0] + g_t32[24][1] * EEEO[1] + add) >> shift);
        for (k = 4; k < 32; k += 8)
        {
            dst[k * line] = (int16_t)((g_t32[k][0] * EEO[0] + g_t32[k][1] * EEO[1] + g_t32[k][2] * EEO[2] +
                                       g_t32[k][3] * EEO[3] + add) >> shift);
        }

        for (k = 2; k < 32; k += 4)
        {
            dst[k * line] = (int16_t)((g_t32[k][0] * EO[0] + g_t32[k][1] * EO[1] + g_t32[k][2] * EO[2] +
                                       g_t32[k][3] * EO[3] + g_t32[k][4] * EO[4] + g_t32[k][5] * EO[5] +
                                       g_t32[k][6] * EO[6] + g_t32[k][7] * EO[7] + add) >> shift);
        }

        for (k = 1; k < 32; k += 2)
        {
            dst[k * line] = (int16_t)((g_t32[k][0] * O[0] + g_t32[k][1] * O[1] + g_t32[k][2] * O[2] + g_t32[k][3] * O[3] +
                                       g_t32[k][4] * O[4] + g_t32[k][5] * O[5] + g_t32[k][6] * O[6] + g_t32[k][7] * O[7] +
                                       g_t32[k][8] * O[8] + g_t32[k][9] * O[9] + g_t32[k][10] * O[10] + g_t32[k][11] *
                                       O[11] + g_t32[k][12] * O[12] + g_t32[k][13] * O[13] + g_t32[k][14] * O[14] +
                                       g_t32[k][15] * O[15] + add) >> shift);
        }

        src += 32;
        dst++;
    }
}

g_t4，g_t8，g_t16，g_t32是几个dct变换的矩阵的系数。

partialButterfly4是块大小为4*4的dct蝶形变换

partialButterfly8是块大小为8*8的dct蝶形变换

partialButterfly16是块大小为16*16的dct蝶形变换

partialButterfly32是块大小为32*32的dct蝶形变换

 

可以通过先调用一次partialButterfly16对行做dct变换，在通过partialButterfly16对列做dct变换

static void dct4_c(const int16_t* src, int16_t* dst, intptr_t srcStride)
{
    const int shift_1st = 1 + X265_DEPTH - 8;
    const int shift_2nd = 8;

    ALIGN_VAR_32(int16_t, coef[4 * 4]);
    ALIGN_VAR_32(int16_t, block[4 * 4]);

    for (int i = 0; i < 4; i++)
    {
        memcpy(&block[i * 4], &src[i * srcStride], 4 * sizeof(int16_t));
    }

    partialButterfly4(block, coef, shift_1st, 4);
    partialButterfly4(coef, dst, shift_2nd, 4);
}

static void dct8_c(const int16_t* src, int16_t* dst, intptr_t srcStride)
{
    const int shift_1st = 2 + X265_DEPTH - 8;
    const int shift_2nd = 9;

    ALIGN_VAR_32(int16_t, coef[8 * 8]);
    ALIGN_VAR_32(int16_t, block[8 * 8]);

    for (int i = 0; i < 8; i++)
    {
        memcpy(&block[i * 8], &src[i * srcStride], 8 * sizeof(int16_t));
    }

    partialButterfly8(block, coef, shift_1st, 8);
    partialButterfly8(coef, dst, shift_2nd, 8);
}

static void dct16_c(const int16_t* src, int16_t* dst, intptr_t srcStride)
{
    const int shift_1st = 3 + X265_DEPTH - 8;
    const int shift_2nd = 10;

    ALIGN_VAR_32(int16_t, coef[16 * 16]);
    ALIGN_VAR_32(int16_t, block[16 * 16]);

    for (int i = 0; i < 16; i++)
    {
        memcpy(&block[i * 16], &src[i * srcStride], 16 * sizeof(int16_t));
    }

    partialButterfly16(block, coef, shift_1st, 16);
    partialButterfly16(coef, dst, shift_2nd, 16);
}

static void dct32_c(const int16_t* src, int16_t* dst, intptr_t srcStride)
{
    const int shift_1st = 4 + X265_DEPTH - 8;
    const int shift_2nd = 11;

    ALIGN_VAR_32(int16_t, coef[32 * 32]);
    ALIGN_VAR_32(int16_t, block[32 * 32]);

    for (int i = 0; i < 32; i++)
    {
        memcpy(&block[i * 32], &src[i * srcStride], 32 * sizeof(int16_t));
    }

    partialButterfly32(block, coef, shift_1st, 32);
    partialButterfly32(coef, dst, shift_2nd, 32);
}

以下是idct变换，调用方式和dct变换类似

static void partialButterflyInverse4(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j;
    int E[2], O[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */
        O[0] = g_t4[1][0] * src[line] + g_t4[3][0] * src[3 * line];
        O[1] = g_t4[1][1] * src[line] + g_t4[3][1] * src[3 * line];
        E[0] = g_t4[0][0] * src[0] + g_t4[2][0] * src[2 * line];
        E[1] = g_t4[0][1] * src[0] + g_t4[2][1] * src[2 * line];

        /* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
        dst[0] = (int16_t)(x265_clip3(-32768, 32767, (E[0] + O[0] + add) >> shift));
        dst[1] = (int16_t)(x265_clip3(-32768, 32767, (E[1] + O[1] + add) >> shift));
        dst[2] = (int16_t)(x265_clip3(-32768, 32767, (E[1] - O[1] + add) >> shift));
        dst[3] = (int16_t)(x265_clip3(-32768, 32767, (E[0] - O[0] + add) >> shift));

        src++;
        dst += 4;
    }
}

static void partialButterflyInverse8(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[4], O[4];
    int EE[2], EO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */
        for (k = 0; k < 4; k++)
        {
            O[k] = g_t8[1][k] * src[line] + g_t8[3][k] * src[3 * line] + g_t8[5][k] * src[5 * line] + g_t8[7][k] * src[7 * line];
        }

        EO[0] = g_t8[2][0] * src[2 * line] + g_t8[6][0] * src[6 * line];
        EO[1] = g_t8[2][1] * src[2 * line] + g_t8[6][1] * src[6 * line];
        EE[0] = g_t8[0][0] * src[0] + g_t8[4][0] * src[4 * line];
        EE[1] = g_t8[0][1] * src[0] + g_t8[4][1] * src[4 * line];

        /* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
        E[0] = EE[0] + EO[0];
        E[3] = EE[0] - EO[0];
        E[1] = EE[1] + EO[1];
        E[2] = EE[1] - EO[1];
        for (k = 0; k < 4; k++)
        {
            dst[k] = (int16_t)x265_clip3(-32768, 32767, (E[k] + O[k] + add) >> shift);
            dst[k + 4] = (int16_t)x265_clip3(-32768, 32767, (E[3 - k] - O[3 - k] + add) >> shift);
        }

        src++;
        dst += 8;
    }
}

static void partialButterflyInverse16(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[8], O[8];
    int EE[4], EO[4];
    int EEE[2], EEO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */
        for (k = 0; k < 8; k++)
        {
            O[k] = g_t16[1][k] * src[line] + g_t16[3][k] * src[3 * line] + g_t16[5][k] * src[5 * line] + g_t16[7][k] * src[7 * line] +
                g_t16[9][k] * src[9 * line] + g_t16[11][k] * src[11 * line] + g_t16[13][k] * src[13 * line] + g_t16[15][k] * src[15 * line];
        }

        for (k = 0; k < 4; k++)
        {
            EO[k] = g_t16[2][k] * src[2 * line] + g_t16[6][k] * src[6 * line] + g_t16[10][k] * src[10 * line] + g_t16[14][k] * src[14 * line];
        }

        EEO[0] = g_t16[4][0] * src[4 * line] + g_t16[12][0] * src[12 * line];
        EEE[0] = g_t16[0][0] * src[0] + g_t16[8][0] * src[8 * line];
        EEO[1] = g_t16[4][1] * src[4 * line] + g_t16[12][1] * src[12 * line];
        EEE[1] = g_t16[0][1] * src[0] + g_t16[8][1] * src[8 * line];

        /* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
        for (k = 0; k < 2; k++)
        {
            EE[k] = EEE[k] + EEO[k];
            EE[k + 2] = EEE[1 - k] - EEO[1 - k];
        }

        for (k = 0; k < 4; k++)
        {
            E[k] = EE[k] + EO[k];
            E[k + 4] = EE[3 - k] - EO[3 - k];
        }

        for (k = 0; k < 8; k++)
        {
            dst[k]   = (int16_t)x265_clip3(-32768, 32767, (E[k] + O[k] + add) >> shift);
            dst[k + 8] = (int16_t)x265_clip3(-32768, 32767, (E[7 - k] - O[7 - k] + add) >> shift);
        }

        src++;
        dst += 16;
    }
}

static void partialButterflyInverse32(const int16_t* src, int16_t* dst, int shift, int line)
{
    int j, k;
    int E[16], O[16];
    int EE[8], EO[8];
    int EEE[4], EEO[4];
    int EEEE[2], EEEO[2];
    int add = 1 << (shift - 1);

    for (j = 0; j < line; j++)
    {
        /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */
        for (k = 0; k < 16; k++)
        {
            O[k] = g_t32[1][k] * src[line] + g_t32[3][k] * src[3 * line] + g_t32[5][k] * src[5 * line] + g_t32[7][k] * src[7 * line] +
                g_t32[9][k] * src[9 * line] + g_t32[11][k] * src[11 * line] + g_t32[13][k] * src[13 * line] + g_t32[15][k] * src[15 * line] +
                g_t32[17][k] * src[17 * line] + g_t32[19][k] * src[19 * line] + g_t32[21][k] * src[21 * line] + g_t32[23][k] * src[23 * line] +
                g_t32[25][k] * src[25 * line] + g_t32[27][k] * src[27 * line] + g_t32[29][k] * src[29 * line] + g_t32[31][k] * src[31 * line];
        }

        for (k = 0; k < 8; k++)
        {
            EO[k] = g_t32[2][k] * src[2 * line] + g_t32[6][k] * src[6 * line] + g_t32[10][k] * src[10 * line] + g_t32[14][k] * src[14 * line] +
                g_t32[18][k] * src[18 * line] + g_t32[22][k] * src[22 * line] + g_t32[26][k] * src[26 * line] + g_t32[30][k] * src[30 * line];
        }

        for (k = 0; k < 4; k++)
        {
            EEO[k] = g_t32[4][k] * src[4 * line] + g_t32[12][k] * src[12 * line] + g_t32[20][k] * src[20 * line] + g_t32[28][k] * src[28 * line];
        }

        EEEO[0] = g_t32[8][0] * src[8 * line] + g_t32[24][0] * src[24 * line];
        EEEO[1] = g_t32[8][1] * src[8 * line] + g_t32[24][1] * src[24 * line];
        EEEE[0] = g_t32[0][0] * src[0] + g_t32[16][0] * src[16 * line];
        EEEE[1] = g_t32[0][1] * src[0] + g_t32[16][1] * src[16 * line];

        /* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
        EEE[0] = EEEE[0] + EEEO[0];
        EEE[3] = EEEE[0] - EEEO[0];
        EEE[1] = EEEE[1] + EEEO[1];
        EEE[2] = EEEE[1] - EEEO[1];
        for (k = 0; k < 4; k++)
        {
            EE[k] = EEE[k] + EEO[k];
            EE[k + 4] = EEE[3 - k] - EEO[3 - k];
        }

        for (k = 0; k < 8; k++)
        {
            E[k] = EE[k] + EO[k];
            E[k + 8] = EE[7 - k] - EO[7 - k];
        }

        for (k = 0; k < 16; k++)
        {
            dst[k] = (int16_t)x265_clip3(-32768, 32767, (E[k] + O[k] + add) >> shift);
            dst[k + 16] = (int16_t)x265_clip3(-32768, 32767, (E[15 - k] - O[15 - k] + add) >> shift);
        }

        src++;
        dst += 32;
    }
}

 

static void idst4_c(const int16_t* src, int16_t* dst, intptr_t dstStride)
{
    const int shift_1st = 7;
    const int shift_2nd = 12 - (X265_DEPTH - 8);

    ALIGN_VAR_32(int16_t, coef[4 * 4]);
    ALIGN_VAR_32(int16_t, block[4 * 4]);

    inversedst(src, coef, shift_1st); // Forward DST BY FAST ALGORITHM, block input, coef output
    inversedst(coef, block, shift_2nd); // Forward DST BY FAST ALGORITHM, coef input, coeff output

    for (int i = 0; i < 4; i++)
    {
        memcpy(&dst[i * dstStride], &block[i * 4], 4 * sizeof(int16_t));
    }
}

static void idct4_c(const int16_t* src, int16_t* dst, intptr_t dstStride)
{
    const int shift_1st = 7;
    const int shift_2nd = 12 - (X265_DEPTH - 8);

    ALIGN_VAR_32(int16_t, coef[4 * 4]);
    ALIGN_VAR_32(int16_t, block[4 * 4]);

    partialButterflyInverse4(src, coef, shift_1st, 4); // Forward DST BY FAST ALGORITHM, block input, coef output
    partialButterflyInverse4(coef, block, shift_2nd, 4); // Forward DST BY FAST ALGORITHM, coef input, coeff output

    for (int i = 0; i < 4; i++)
    {
        memcpy(&dst[i * dstStride], &block[i * 4], 4 * sizeof(int16_t));
    }
}

static void idct8_c(const int16_t* src, int16_t* dst, intptr_t dstStride)
{
    const int shift_1st = 7;
    const int shift_2nd = 12 - (X265_DEPTH - 8);

    ALIGN_VAR_32(int16_t, coef[8 * 8]);
    ALIGN_VAR_32(int16_t, block[8 * 8]);

    partialButterflyInverse8(src, coef, shift_1st, 8);
    partialButterflyInverse8(coef, block, shift_2nd, 8);

    for (int i = 0; i < 8; i++)
    {
        memcpy(&dst[i * dstStride], &block[i * 8], 8 * sizeof(int16_t));
    }
}

static void idct16_c(const int16_t* src, int16_t* dst, intptr_t dstStride)
{
    const int shift_1st = 7;
    const int shift_2nd = 12 - (X265_DEPTH - 8);

    ALIGN_VAR_32(int16_t, coef[16 * 16]);
    ALIGN_VAR_32(int16_t, block[16 * 16]);

    partialButterflyInverse16(src, coef, shift_1st, 16);
    partialButterflyInverse16(coef, block, shift_2nd, 16);

    for (int i = 0; i < 16; i++)
    {
        memcpy(&dst[i * dstStride], &block[i * 16], 16 * sizeof(int16_t));
    }
}

static void idct32_c(const int16_t* src, int16_t* dst, intptr_t dstStride)
{
    const int shift_1st = 7;
    const int shift_2nd = 12 - (X265_DEPTH - 8);

    ALIGN_VAR_32(int16_t, coef[32 * 32]);
    ALIGN_VAR_32(int16_t, block[32 * 32]);

    partialButterflyInverse32(src, coef, shift_1st, 32);
    partialButterflyInverse32(coef, block, shift_2nd, 32);

    for (int i = 0; i < 32; i++)
    {
        memcpy(&dst[i * dstStride], &block[i * 32], 32 * sizeof(int16_t));
    }
}

 

被挫伤的雨
关注
  • 0
    点赞
  • 踩
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
【Codecs系列】HEVC标准（七）：变换量化技术及x265实现分析
CSDN明星博主，认证博客专家，视频、Matlab领域优质创作者。
01-05 1112
Date: 2022.1.5 文章目录1、变换编码1.1、整数DCT1.1.1、基本原理1.1.2、公式推导过程1.2、整数DST1.3、DCT反变换1.4、变换四叉树1.5、哈达玛变换2、量化技术2.1、传统标量量化2.2、反量化2.3、RDOQ2.4、量化参数和量化矩阵 1、变换编码 图像变换编码是指将以空间域中像素形式描述的图像转换至空间域，以变换系数的形式加以表示。绝大多数图像都含有较多平坦区域和内容变换缓慢的区域，适当的变换可使图像能量在空间域的分散分布转为在变换域的相对集中分布，以达到去除空间
图像压缩编码和解码原理
tianyu的专栏 - Linux site:blog.csdn.net/wishfly
07-29 2万+
本节介绍图像压缩编码的基本原理，图像数据压缩和解压缩电路的基本结构。它们是看影碟机电路图的基础知识。 　　一、图像压缩的基本途径 　　图像的数据量极大，必须对其数据总量大大压缩，才能够存储在直径12cm的光盘上。在实用技术上，可通过以下途径来压缩图像数据的总量。 　　1、采用亮度(Y)、色度(C)取样方式 　　实用彩色电视技术没有传输、处理红、蓝、绿三基色信号，而传输、处理
使用可变块大小模拟 DCT、Walsh、Hadamard、Haar 和 Slant 变换：使用 dct 矩阵对灰度图像和 DCT 执行非正弦图像变换。-matlab开发
05-30
使用给定块大小顺序的 dct 矩阵对灰度图像和 DCT 执行非正弦图像变换。 重建的图像与描述变换是无损的原始图像完全相同。 此外，转换中还包含了生成 walsh、slant 和 haar 矩阵的代码。
DCT函数类型
weixin_39087379的博客
05-21 192
【DCT笔记】DCT变换、DCT反变换、分块DCT变换
weixin_30609331的博客
08-31 9097
DCT变换、DCT反变换、分块DCT变换 欢迎转载，但请注明出处！ 一、引言 DCT变换的全称是离散余弦变换(Discrete Cosine Transform)，主要用于将数据或图像的压缩，能够将空域的信号转换到频域上，具有良好的去相关性的性能。DCT变换本身是无损的，但是在图像编码等领域给接下来的量化、哈弗曼编码等创造了很好的条件，同时，由于DCT变换时对称的，所...
基于FPGA的DCT变换verilog设计+word版论文。
05-14
首先对整数DCT变换常见快速算法进行了分析，得出了基于类DCT矩阵进行变换的DCT变换较其它的快速算法更适合硬件实现，并利用Verilog HDL硬件描述语言开发了RTL级代码，并进行了功能仿真，仿真结果证实了设计的正确性...
IMG_DCT变换矩阵算法_
09-30
提高压缩比，DCT快速变换的限制较少，该转换技术只需要在实数域上就可以实现
基于DCT变换的数字水印算法
11-21
DCT离散余弦变换实现数字水印算法；在进行水印嵌入时，先对水印用了Arnold变换；包含了几种不同的水印攻击
dct.rar_DCT变换
09-22
dct变换 先进fft变换，然后再利用fft进行dct变换，是基于win32平台的
基于DCT变换的图像量化matlab仿真
最新发布
01-03
在DCT变换后，需要对得到的DCT系数进行量化处理。量化是将连续的数值映射为有限的离散值的过程。在图像压缩中，量化可以减少图像数据的存储和传输需求。具体来说，对于DCT系数，可以采用标量量化或矢量量化的方法，...
DCT变换(离散余弦变换)的矩阵乘法实现(64×64维)
06-18
DCT变换(离散余弦变换)的矩阵乘法实现(64×64维)
HEVC DCT变换
qq_35752161的博客
01-14 675
注意，由于量化和反量化可能带来数值溢出，行反变换多右移1位 右移的存在精度补偿 DCT正变换 Void xTrMxN(Int bitDepth, TCoeff *block, TCoeff *coeff, Int iWidth, Int iHeight, Bool useDST, const Int maxLog2TrDynamicRange) { const Int TRANSFORM_MATRIX_SHIFT = g_transformMatrixShi...
HEVC学习（十） —— 与变换有关的几个主要函数及重要变量
HEVC_CJL的专栏
11-30 9019
本文主要是列出几个与变换有关的函数，以及涉及到的比较重要的变量，但不会去深入讨论函数的实现，日后有机会的话会更详细地讨论这几个函数，这里仅是列举出来，起到备忘的作用，也能供大家参考。   如何定位到这几个函数的，方法与前面的一篇博客HEVC学习（八） —— 以SAO为例浅析跟踪代码方法类似，有兴趣的朋友可以去看看，这里就不重复这个过程了。   在xCompressCU函数中，有这么几个函数
离散余弦变换（DCT）的来龙去脉
热门推荐
dugudaibo的博客
11-01 5万+
本文主要介绍离散余弦变换（DCT），从图像的二维离散变换入手，引出变换核的可分性与对称性导出DCT，并给出实现代码。
HEVC函数入门（8）——变换的实现
不是他人太努力，只是自己努力不够罢了
06-21 1849
前面我们讲的都是预测，下面开始变换的实现。 本文参考http://blog.csdn.net/shaqoneal/article/details/44856469 1、HM中Intra模式的主要实现逻辑 以Intra的亮度模式为例。主要实现代码实现于TEncSearch::estIntraPredQT方法中。TEncSearch::estIntraPredQT实现时，首先获取当前CU的分割子块
HEVC代码学习36：xTrMxN函数
慢慢积累
01-15 939
之前在transformNxN的学习中提到，xT函数是前向变换的入口函数，实际上xT完成的是变换前的准备工作，真正的前向变换由xTrMxN函数实现。 先来看下xT函数。xT只是一个入口函数，首先定义了残差块和系数块，然后调用xTrMxN进行前向变换。 /** Wrapper function between HM interface and core NxN forward transfo
H265和H264编解码中DCT变换过程中scaling 和 strong_intra_smooth
cy413026的博客
06-26 1294
strong_intra_smooth:是H265的intra预测过程中，对相邻pix进行的双线性滤波，属于协议内容 scaling是DCT变换过程中的，用于归一化，为了保证残差块经过变换和反变换之后基准相同，HEVC在变换模块中还增加了额外的缩放参考HEVC中变换（Transform）过程中的scaling操作的理解 ...
X265的编码流程
dragon_dy的博客
09-08 4633
X265的编码主函数 int Encoder::encode(const x265_picture* pic_in, x265_picture* pic_out) 过程为： 1. 先将输入图像x265_picture* pic_in中的数据复制到内部帧结构Frame *inFrame，计算好其它帧信息 2. 调用m_lookahead->addPicture(*inFrame, slice
H.266代码学习：xT，xTrMxN，xTrMxN_EMT函数
慢慢积累
03-13 1004
今天来学习xT和xTrMxN和xTrMxN_EMT函数。 在之前的 H.266代码学习：transformNxN函数 中提到JEM变换分为两次：主变换+二次NSST。transformNxN会调用xT进行主变换。 在JEM中，主变换有三种： 1.原HEVC中的DCT-2/4x4 DST。 2.多核变换。 3.KLT变换。 其中1和3的入口函数为xTrMxN，2的入口函数为xTrMxN_...
fpga图像压缩dct
08-11
FPGA（Field-Programmable Gate Array）图像压缩算法中的DCT（离散余弦变换）是一种常用的图像压缩技术。DCT通过将图像分解成若干频率分量来实现压缩。 在FPGA中实现DCT图像压缩算法有几个关键步骤。首先，将输入图像分割成若干大小相等的图像块。对于每个图像块，进行预处理，如减去均值，以消除直流分量。然后，对每个图像块应用离散余弦变换，将图像从时域转换到频域。 在DCT变换之后，通常会对高频系数进行量化，以减少数据的维度和冗余。然后，使用霍夫曼编码或其他编码方法对量化后的系数进行熵编码，以实现高效压缩。 在FPGA中实现DCT图像压缩算法可以提供高性能和低功耗的优势。由于FPGA可编程的特性，可以对算法进行硬件加速，实现并行计算，从而提高处理速度。此外，FPGA还可以根据需求进行灵活的优化和定制，以满足不同应用的需求。 总的来说，FPGA图像压缩中的DCT算法通过将图像转换到频域，并进行量化和熵编码来实现压缩。FPGA提供了高性能和低功耗的计算平台，可以针对特定应用进行优化，从而更好地满足对图像压缩的需求。

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值