x264代码剖析(十六):核心算法之宏块编码中的量化编码

简介: <h1 style="text-align:center"><strong>x264<span style="font-family:宋体">代码剖析(十六):核心算法之宏块编码中的量化编码</span></strong></h1> <p> </p> <p>        为了进一步节省图像的传输码率,需要对图像进行压缩,通常采用变换编码及量化来消除图像中的相关性以减少图像编码的动态范

x264代码剖析(十六):核心算法之宏块编码中的量化编码

 

        为了进一步节省图像的传输码率,需要对图像进行压缩,通常采用变换编码及量化来消除图像中的相关性以减少图像编码的动态范围。本文主要介绍量化的相关内容,并给出x264中量化编码的代码分析。

 

1、量化编码

 

        量化过程就是根据图像的动态范围大小确定量化参数,既保留图像必要的细节,又可以减少码流。在图像编码中,变换编码和量化从原理上讲是两个独立的过程。但在H.264中,将两个过程中的乘法合二为一,并进一步采用整数运算,减少编解码的运算量,提高图像压缩的实时性,这些措施对峰值信噪比(PSNR)的影响很小,一般低于0.02dB,可不计。下面给出编码器中变换编码及量化过程的流程。




        下面给出毕厚杰著作的《新一代视频压缩编码标准——H.264/AVC(第二版)》中量化的相关理论知识,仅供学习参考。




2、量化源码分析

 

        量化的相关函数大都位于common/quant.c中,主要包括以下函数:

x264_quant_init():初始化量化和反量化相关的汇编函数

quant_4x4():对4x4DCT残差矩阵进行量化

QUANT_ONE():完成一个DCT系数的量化工作

quant_4x4x4():对44x4DCT残差矩阵进行量化

 

        下面对这些函数进行深入分析。


2.1x264_quant_init()函数

 

        量化模块的初始化函数是x264_quant_init()。该函数对x264_quant_function_t结构体中的函数指针进行了赋值。X264运行的过程中只要调用x264_quant_function_t的函数指针就可以完成相应的功能。对应的代码分析如下:

 

/******************************************************************/
/******************************************************************/
/*
======Analysed by RuiDong Fang
======Csdn Blog:http://blog.csdn.net/frd2009041510
======Date:2016.03.25
 */
/******************************************************************/
/******************************************************************/

/************====== 量化 ======************/
/*
功能:量化初始化,x264_quant_init()被x264_encoder_open()函数调用
*/
void x264_quant_init( x264_t *h, int cpu, x264_quant_function_t *pf )
{
    /*===== 量化=====*/
	pf->quant_8x8 = quant_8x8;	/////////////////针对8x8DCT的量化
    pf->quant_4x4 = quant_4x4;	/////////////////量化4x4=16个
    pf->quant_4x4x4 = quant_4x4x4;	/////////////////处理4个4x4的块

	/////////////////Intra16x16中,16个DC系数Hadamard变换后对的它们量化
    pf->quant_4x4_dc = quant_4x4_dc;	
    pf->quant_2x2_dc = quant_2x2_dc;

	/*===== 反量化=====*/
    pf->dequant_4x4 = dequant_4x4;	/////////////////反量化4x4=16个
    pf->dequant_4x4_dc = dequant_4x4_dc;	/////////////////
    pf->dequant_8x8 = dequant_8x8;	/////////////////处理4个4x4的块针对8x8DCT的反量化

	/////////////////Intra16x16中,16个DC系数Hadamard变换后对的它们反量化
    pf->idct_dequant_2x4_dc = idct_dequant_2x4_dc;
    pf->idct_dequant_2x4_dconly = idct_dequant_2x4_dconly;

	//色度
    pf->optimize_chroma_2x2_dc = optimize_chroma_2x2_dc;
    pf->optimize_chroma_2x4_dc = optimize_chroma_2x4_dc;

    pf->denoise_dct = x264_denoise_dct;
    pf->decimate_score15 = x264_decimate_score15;
    pf->decimate_score16 = x264_decimate_score16;
    pf->decimate_score64 = x264_decimate_score64;

    pf->coeff_last4 = x264_coeff_last4;
    pf->coeff_last8 = x264_coeff_last8;
    pf->coeff_last[  DCT_LUMA_AC] = x264_coeff_last15;
    pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16;
    pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64;
    pf->coeff_level_run4 = x264_coeff_level_run4;
    pf->coeff_level_run8 = x264_coeff_level_run8;
    pf->coeff_level_run[  DCT_LUMA_AC] = x264_coeff_level_run15;
    pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16;

#if HIGH_BIT_DEPTH
#if HAVE_MMX
    INIT_TRELLIS( sse2 );
    if( cpu&X264_CPU_MMX2 )
    {
#if ARCH_X86
        pf->denoise_dct = x264_denoise_dct_mmx;
        pf->decimate_score15 = x264_decimate_score15_mmx2;
        pf->decimate_score16 = x264_decimate_score16_mmx2;
        pf->decimate_score64 = x264_decimate_score64_mmx2;
        pf->coeff_last8 = x264_coeff_last8_mmx2;
        pf->coeff_last[  DCT_LUMA_AC] = x264_coeff_last15_mmx2;
        pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16_mmx2;
        pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64_mmx2;
        pf->coeff_level_run8 = x264_coeff_level_run8_mmx2;
        pf->coeff_level_run[  DCT_LUMA_AC] = x264_coeff_level_run15_mmx2;
        pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16_mmx2;
#endif
        pf->coeff_last4 = x264_coeff_last4_mmx2;
        pf->coeff_level_run4 = x264_coeff_level_run4_mmx2;
        if( cpu&X264_CPU_LZCNT )
            pf->coeff_level_run4 = x264_coeff_level_run4_mmx2_lzcnt;
    }
    if( cpu&X264_CPU_SSE2 )
    {
        pf->quant_4x4 = x264_quant_4x4_sse2;
        pf->quant_4x4x4 = x264_quant_4x4x4_sse2;
        pf->quant_8x8 = x264_quant_8x8_sse2;
        pf->quant_2x2_dc = x264_quant_2x2_dc_sse2;
        pf->quant_4x4_dc = x264_quant_4x4_dc_sse2;
        pf->dequant_4x4 = x264_dequant_4x4_sse2;
        pf->dequant_8x8 = x264_dequant_8x8_sse2;
        pf->dequant_4x4_dc = x264_dequant_4x4dc_sse2;
        pf->denoise_dct = x264_denoise_dct_sse2;
        pf->decimate_score15 = x264_decimate_score15_sse2;
        pf->decimate_score16 = x264_decimate_score16_sse2;
        pf->decimate_score64 = x264_decimate_score64_sse2;
        pf->coeff_last8 = x264_coeff_last8_sse2;
        pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2;
        pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2;
        pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2;
        pf->coeff_level_run8 = x264_coeff_level_run8_sse2;
        pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2;
        pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2;
        if( cpu&X264_CPU_LZCNT )
        {
            pf->coeff_last4 = x264_coeff_last4_mmx2_lzcnt;
            pf->coeff_last8 = x264_coeff_last8_sse2_lzcnt;
            pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2_lzcnt;
            pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2_lzcnt;
            pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2_lzcnt;
            pf->coeff_level_run8 = x264_coeff_level_run8_sse2_lzcnt;
            pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2_lzcnt;
            pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2_lzcnt;
        }
    }
    if( cpu&X264_CPU_SSSE3 )
    {
        pf->quant_4x4 = x264_quant_4x4_ssse3;
        pf->quant_4x4x4 = x264_quant_4x4x4_ssse3;
        pf->quant_8x8 = x264_quant_8x8_ssse3;
        pf->quant_2x2_dc = x264_quant_2x2_dc_ssse3;
        pf->quant_4x4_dc = x264_quant_4x4_dc_ssse3;
        pf->denoise_dct = x264_denoise_dct_ssse3;
        pf->decimate_score15 = x264_decimate_score15_ssse3;
        pf->decimate_score16 = x264_decimate_score16_ssse3;
        pf->decimate_score64 = x264_decimate_score64_ssse3;
        INIT_TRELLIS( ssse3 );
    }
    if( cpu&X264_CPU_SSE4 )
    {
        pf->quant_2x2_dc = x264_quant_2x2_dc_sse4;
        pf->quant_4x4_dc = x264_quant_4x4_dc_sse4;
        pf->quant_4x4 = x264_quant_4x4_sse4;
        pf->quant_4x4x4 = x264_quant_4x4x4_sse4;
        pf->quant_8x8 = x264_quant_8x8_sse4;
    }
    if( cpu&X264_CPU_AVX )
    {
        pf->denoise_dct = x264_denoise_dct_avx;
    }
    if( cpu&X264_CPU_XOP )
    {
        pf->dequant_4x4_dc = x264_dequant_4x4dc_xop;
        if( h->param.i_cqm_preset != X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_xop;
            pf->dequant_8x8 = x264_dequant_8x8_xop;
        }
    }
    if( cpu&X264_CPU_AVX2 )
    {
        pf->quant_4x4 = x264_quant_4x4_avx2;
        pf->quant_4x4_dc = x264_quant_4x4_dc_avx2;
        pf->quant_8x8 = x264_quant_8x8_avx2;
        pf->quant_4x4x4 = x264_quant_4x4x4_avx2;
        pf->dequant_4x4 = x264_dequant_4x4_avx2;
        pf->dequant_8x8 = x264_dequant_8x8_avx2;
        pf->dequant_4x4_dc = x264_dequant_4x4dc_avx2;
        pf->denoise_dct = x264_denoise_dct_avx2;
        if( cpu&X264_CPU_LZCNT )
            pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_avx2_lzcnt;
    }
#endif // HAVE_MMX
#else // !HIGH_BIT_DEPTH
#if HAVE_MMX
    INIT_TRELLIS( sse2 );
    if( cpu&X264_CPU_MMX )
    {
#if ARCH_X86
        pf->dequant_4x4 = x264_dequant_4x4_mmx;
        pf->dequant_4x4_dc = x264_dequant_4x4dc_mmx2;
        pf->dequant_8x8 = x264_dequant_8x8_mmx;
        if( h->param.i_cqm_preset == X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_flat16_mmx;
            pf->dequant_8x8 = x264_dequant_8x8_flat16_mmx;
        }
        pf->denoise_dct = x264_denoise_dct_mmx;
#endif
    }

    if( cpu&X264_CPU_MMX2 )
    {
        pf->quant_2x2_dc = x264_quant_2x2_dc_mmx2;
#if ARCH_X86
        pf->quant_4x4 = x264_quant_4x4_mmx2;
        pf->quant_8x8 = x264_quant_8x8_mmx2;
        pf->quant_4x4_dc = x264_quant_4x4_dc_mmx2;
        pf->decimate_score15 = x264_decimate_score15_mmx2;
        pf->decimate_score16 = x264_decimate_score16_mmx2;
        pf->decimate_score64 = x264_decimate_score64_mmx2;
        pf->coeff_last[  DCT_LUMA_AC] = x264_coeff_last15_mmx2;
        pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16_mmx2;
        pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64_mmx2;
        pf->coeff_level_run[  DCT_LUMA_AC] = x264_coeff_level_run15_mmx2;
        pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16_mmx2;
#endif
        pf->coeff_last4 = x264_coeff_last4_mmx2;
        pf->coeff_last8 = x264_coeff_last8_mmx2;
        pf->coeff_level_run4 = x264_coeff_level_run4_mmx2;
        pf->coeff_level_run8 = x264_coeff_level_run8_mmx2;
        if( cpu&X264_CPU_LZCNT )
        {
            pf->coeff_last4 = x264_coeff_last4_mmx2_lzcnt;
            pf->coeff_last8 = x264_coeff_last8_mmx2_lzcnt;
            pf->coeff_level_run4 = x264_coeff_level_run4_mmx2_lzcnt;
            pf->coeff_level_run8 = x264_coeff_level_run8_mmx2_lzcnt;
        }
    }

    if( cpu&X264_CPU_SSE2 )
    {
        pf->quant_4x4_dc = x264_quant_4x4_dc_sse2;
        pf->quant_4x4 = x264_quant_4x4_sse2;
        pf->quant_4x4x4 = x264_quant_4x4x4_sse2;
        pf->quant_8x8 = x264_quant_8x8_sse2;
        pf->dequant_4x4 = x264_dequant_4x4_sse2;
        pf->dequant_4x4_dc = x264_dequant_4x4dc_sse2;
        pf->dequant_8x8 = x264_dequant_8x8_sse2;
        if( h->param.i_cqm_preset == X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_flat16_sse2;
            pf->dequant_8x8 = x264_dequant_8x8_flat16_sse2;
        }
        pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_sse2;
        pf->denoise_dct = x264_denoise_dct_sse2;
        pf->decimate_score15 = x264_decimate_score15_sse2;
        pf->decimate_score16 = x264_decimate_score16_sse2;
        pf->decimate_score64 = x264_decimate_score64_sse2;
        pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2;
        pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2;
        pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2;
        pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2;
        pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2;
        if( cpu&X264_CPU_LZCNT )
        {
            pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2_lzcnt;
            pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2_lzcnt;
            pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2_lzcnt;
            pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2_lzcnt;
            pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2_lzcnt;
        }
    }

    if( cpu&X264_CPU_SSSE3 )
    {
        pf->quant_2x2_dc = x264_quant_2x2_dc_ssse3;
        pf->quant_4x4_dc = x264_quant_4x4_dc_ssse3;
        pf->quant_4x4 = x264_quant_4x4_ssse3;
        pf->quant_4x4x4 = x264_quant_4x4x4_ssse3;
        pf->quant_8x8 = x264_quant_8x8_ssse3;
        pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_ssse3;
        pf->denoise_dct = x264_denoise_dct_ssse3;
        pf->decimate_score15 = x264_decimate_score15_ssse3;
        pf->decimate_score16 = x264_decimate_score16_ssse3;
        pf->decimate_score64 = x264_decimate_score64_ssse3;
        INIT_TRELLIS( ssse3 );
        pf->coeff_level_run4 = x264_coeff_level_run4_ssse3;
        pf->coeff_level_run8 = x264_coeff_level_run8_ssse3;
        pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_ssse3;
        pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_ssse3;
        if( cpu&X264_CPU_LZCNT )
        {
            pf->coeff_level_run4 = x264_coeff_level_run4_ssse3;
            pf->coeff_level_run8 = x264_coeff_level_run8_ssse3;
            pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_ssse3_lzcnt;
            pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_ssse3_lzcnt;
        }
    }

    if( cpu&X264_CPU_SSE4 )
    {
        pf->quant_4x4_dc = x264_quant_4x4_dc_sse4;
        pf->quant_4x4 = x264_quant_4x4_sse4;
        pf->quant_8x8 = x264_quant_8x8_sse4;
        pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_sse4;
    }

    if( cpu&X264_CPU_AVX )
    {
        pf->dequant_4x4_dc = x264_dequant_4x4dc_avx;
        if( h->param.i_cqm_preset != X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_avx;
            pf->dequant_8x8 = x264_dequant_8x8_avx;
        }
        pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_avx;
        pf->denoise_dct = x264_denoise_dct_avx;
    }

    if( cpu&X264_CPU_XOP )
    {
        if( h->param.i_cqm_preset != X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_xop;
            pf->dequant_8x8 = x264_dequant_8x8_xop;
        }
    }

    if( cpu&X264_CPU_AVX2 )
    {
        pf->quant_4x4 = x264_quant_4x4_avx2;
        pf->quant_4x4_dc = x264_quant_4x4_dc_avx2;
        pf->quant_8x8 = x264_quant_8x8_avx2;
        pf->quant_4x4x4 = x264_quant_4x4x4_avx2;
        pf->dequant_4x4 = x264_dequant_4x4_avx2;
        pf->dequant_8x8 = x264_dequant_8x8_avx2;
        pf->dequant_4x4_dc = x264_dequant_4x4dc_avx2;
        if( h->param.i_cqm_preset == X264_CQM_FLAT )
        {
            pf->dequant_4x4 = x264_dequant_4x4_flat16_avx2;
            pf->dequant_8x8 = x264_dequant_8x8_flat16_avx2;
        }
        pf->decimate_score64 = x264_decimate_score64_avx2;
        pf->denoise_dct = x264_denoise_dct_avx2;
        if( cpu&X264_CPU_LZCNT )
        {
            pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_avx2_lzcnt;
            pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_avx2_lzcnt;
            pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_avx2_lzcnt;
        }
    }
#endif // HAVE_MMX

#if HAVE_ALTIVEC
    if( cpu&X264_CPU_ALTIVEC )
    {
        pf->quant_2x2_dc = x264_quant_2x2_dc_altivec;
        pf->quant_4x4_dc = x264_quant_4x4_dc_altivec;
        pf->quant_4x4 = x264_quant_4x4_altivec;
        pf->quant_8x8 = x264_quant_8x8_altivec;

        pf->dequant_4x4 = x264_dequant_4x4_altivec;
        pf->dequant_8x8 = x264_dequant_8x8_altivec;
    }
#endif

#if HAVE_ARMV6
    if( cpu&X264_CPU_ARMV6 )
    {
        pf->coeff_last4 = x264_coeff_last4_arm;
        pf->coeff_last8 = x264_coeff_last8_arm;
    }
#endif
#if HAVE_ARMV6 || ARCH_AARCH64
    if( cpu&X264_CPU_NEON )
    {
        pf->quant_2x2_dc   = x264_quant_2x2_dc_neon;
        pf->quant_4x4      = x264_quant_4x4_neon;
        pf->quant_4x4_dc   = x264_quant_4x4_dc_neon;
        pf->quant_4x4x4    = x264_quant_4x4x4_neon;
        pf->quant_8x8      = x264_quant_8x8_neon;
        pf->dequant_4x4    = x264_dequant_4x4_neon;
        pf->dequant_4x4_dc = x264_dequant_4x4_dc_neon;
        pf->dequant_8x8    = x264_dequant_8x8_neon;
        pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_neon;
        pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_neon;
        pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_neon;
        pf->denoise_dct = x264_denoise_dct_neon;
        pf->decimate_score15 = x264_decimate_score15_neon;
        pf->decimate_score16 = x264_decimate_score16_neon;
        pf->decimate_score64 = x264_decimate_score64_neon;
    }
#endif
#if ARCH_AARCH64
    if( cpu&X264_CPU_ARMV8 )
    {
        pf->coeff_last4 = x264_coeff_last4_aarch64;
        pf->coeff_last8 = x264_coeff_last8_aarch64;
        pf->coeff_level_run4 = x264_coeff_level_run4_aarch64;
    }
    if( cpu&X264_CPU_NEON )
    {
        pf->coeff_level_run8 = x264_coeff_level_run8_neon;
        pf->coeff_level_run[  DCT_LUMA_AC] = x264_coeff_level_run15_neon;
        pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16_neon;
    }
#endif

#if HAVE_MSA
    if( cpu&X264_CPU_MSA )
    {
        pf->quant_4x4      = x264_quant_4x4_msa;
        pf->quant_4x4_dc   = x264_quant_4x4_dc_msa;
        pf->quant_4x4x4    = x264_quant_4x4x4_msa;
        pf->quant_8x8      = x264_quant_8x8_msa;
        pf->dequant_4x4    = x264_dequant_4x4_msa;
        pf->dequant_4x4_dc = x264_dequant_4x4_dc_msa;
        pf->dequant_8x8    = x264_dequant_8x8_msa;
        pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_msa;
        pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_msa;
    }
#endif
#endif // HIGH_BIT_DEPTH
    pf->coeff_last[DCT_LUMA_DC]     = pf->coeff_last[DCT_CHROMAU_DC]  = pf->coeff_last[DCT_CHROMAV_DC] =
    pf->coeff_last[DCT_CHROMAU_4x4] = pf->coeff_last[DCT_CHROMAV_4x4] = pf->coeff_last[DCT_LUMA_4x4];
    pf->coeff_last[DCT_CHROMA_AC]   = pf->coeff_last[DCT_CHROMAU_AC]  =
    pf->coeff_last[DCT_CHROMAV_AC]  = pf->coeff_last[DCT_LUMA_AC];
    pf->coeff_last[DCT_CHROMAU_8x8] = pf->coeff_last[DCT_CHROMAV_8x8] = pf->coeff_last[DCT_LUMA_8x8];

    pf->coeff_level_run[DCT_LUMA_DC]     = pf->coeff_level_run[DCT_CHROMAU_DC]  = pf->coeff_level_run[DCT_CHROMAV_DC] =
    pf->coeff_level_run[DCT_CHROMAU_4x4] = pf->coeff_level_run[DCT_CHROMAV_4x4] = pf->coeff_level_run[DCT_LUMA_4x4];
    pf->coeff_level_run[DCT_CHROMA_AC]   = pf->coeff_level_run[DCT_CHROMAU_AC]  =
    pf->coeff_level_run[DCT_CHROMAV_AC]  = pf->coeff_level_run[DCT_LUMA_AC];
}

 

       其中, x264_quant_function_t的定义如下:


typedef struct
{
    int (*quant_8x8)  ( dctcoef dct[64], udctcoef mf[64], udctcoef bias[64] );
    int (*quant_4x4)  ( dctcoef dct[16], udctcoef mf[16], udctcoef bias[16] );
    int (*quant_4x4x4)( dctcoef dct[4][16], udctcoef mf[16], udctcoef bias[16] );
    int (*quant_4x4_dc)( dctcoef dct[16], int mf, int bias );
    int (*quant_2x2_dc)( dctcoef dct[4], int mf, int bias );

    void (*dequant_8x8)( dctcoef dct[64], int dequant_mf[6][64], int i_qp );
    void (*dequant_4x4)( dctcoef dct[16], int dequant_mf[6][16], int i_qp );
    void (*dequant_4x4_dc)( dctcoef dct[16], int dequant_mf[6][16], int i_qp );

    void (*idct_dequant_2x4_dc)( dctcoef dct[8], dctcoef dct4x4[8][16], int dequant_mf[6][16], int i_qp );
    void (*idct_dequant_2x4_dconly)( dctcoef dct[8], int dequant_mf[6][16], int i_qp );

    int (*optimize_chroma_2x2_dc)( dctcoef dct[4], int dequant_mf );
    int (*optimize_chroma_2x4_dc)( dctcoef dct[8], int dequant_mf );

    void (*denoise_dct)( dctcoef *dct, uint32_t *sum, udctcoef *offset, int size );

    int (*decimate_score15)( dctcoef *dct );
    int (*decimate_score16)( dctcoef *dct );
    int (*decimate_score64)( dctcoef *dct );
    int (*coeff_last[14])( dctcoef *dct );
    int (*coeff_last4)( dctcoef *dct );
    int (*coeff_last8)( dctcoef *dct );
    int (*coeff_level_run[13])( dctcoef *dct, x264_run_level_t *runlevel );
    int (*coeff_level_run4)( dctcoef *dct, x264_run_level_t *runlevel );
    int (*coeff_level_run8)( dctcoef *dct, x264_run_level_t *runlevel );

#define TRELLIS_PARAMS const int *unquant_mf, const uint8_t *zigzag, int lambda2,\
                       int last_nnz, dctcoef *coefs, dctcoef *quant_coefs, dctcoef *dct,\
                       uint8_t *cabac_state_sig, uint8_t *cabac_state_last,\
                       uint64_t level_state0, uint16_t level_state1
    int (*trellis_cabac_4x4)( TRELLIS_PARAMS, int b_ac );
    int (*trellis_cabac_8x8)( TRELLIS_PARAMS, int b_interlaced );
    int (*trellis_cabac_4x4_psy)( TRELLIS_PARAMS, int b_ac, dctcoef *fenc_dct, int psy_trellis );
    int (*trellis_cabac_8x8_psy)( TRELLIS_PARAMS, int b_interlaced, dctcoef *fenc_dct, int psy_trellis );
    int (*trellis_cabac_dc)( TRELLIS_PARAMS, int num_coefs );
    int (*trellis_cabac_chroma_422_dc)( TRELLIS_PARAMS );
} x264_quant_function_t;


 

2.2quant_4x4()函数

 

        quant_4x4()用于对4x4DCT残差矩阵进行量化,quant_4x4()循环16次调用了QUANT_ONE()完成了量化工作。并且将DCT系数值,MF值,bias偏移值直接传递给了该宏。对应的代码分析如下:

 

/************====== 4x4的DCT残差矩阵的量化 ======************/
/*
功能:quant_4x4()用于对4x4的DCT残差矩阵进行量化
*/
static int quant_4x4( dctcoef dct[16], udctcoef mf[16], udctcoef bias[16] )//输入输出都是dct[16]
{
    int nz = 0;
    for( int i = 0; i < 16; i++ )//循环16个元素
        QUANT_ONE( dct[i], mf[i], bias[i] );
    return !!nz;
}

 

        其中,QUANT_ONE()完成了一个DCT系数的量化工作,从QUANT_ONE()的定义可以看出,它实现了上文提到的H.264标准中的量化公式。它的定义如下:

 

/************====== QUANT_ONE() ======************/
/*
功能:量化1个元素
*/
#define QUANT_ONE( coef, mf, f ) \
{ \
    if( (coef) > 0 ) \
        (coef) = (f + (coef)) * (mf) >> 16; \
    else \
        (coef) = - ((f - (coef)) * (mf) >> 16); \
    nz |= (coef); \
}

 

2.3quant_4x4x4()函数

 

        quant_4x4x4()用于对44x4DCT残差矩阵进行量化。该函数的定义位于common\quant.c,从quant_4x4x4()的定义可以看出,该函数相当于调用了4quant_4x4()函数。对应的代码分析如下:


/************====== 4个4x4的DCT残差矩阵的量化 ======************/
/*
功能:quant_4x4x4()用于对4个4x4的DCT残差矩阵进行量化,从quant_4x4x4()的定义可以看出,该函数相当于调用了4次quant_4x4()函数。
*/
static int quant_4x4x4( dctcoef dct[4][16], udctcoef mf[16], udctcoef bias[16] )//输入输出都是dct[4][16] 
{
    int nza = 0;
    for( int j = 0; j < 4; j++ )//处理4个
    {
        int nz = 0;
        for( int i = 0; i < 16; i++ )
            QUANT_ONE( dct[j][i], mf[i], bias[i] );//量化
        nza |= (!!nz)<<j;
    }
    return nza;
}


大笑至此,x264中量化的重点函数就分析完毕了。


目录
相关文章
|
1天前
|
机器学习/深度学习 人工智能 自然语言处理
【自然语言处理】TF-IDF算法在人工智能方面的应用,附带代码
TF-IDF算法在人工智能领域,特别是自然语言处理(NLP)和信息检索中,被广泛用于特征提取和文本表示。以下是一个使用Python的scikit-learn库实现TF-IDF算法的简单示例,并展示如何将其应用于文本数据。
113 64
|
1天前
|
机器学习/深度学习 人工智能 算法
【人工智能】传统语音识别算法概述,应用场景,项目实践及案例分析,附带代码示例
传统语音识别算法是将语音信号转化为文本形式的技术,它主要基于模式识别理论和数学统计学方法。以下是传统语音识别算法的基本概述
7 2
|
5天前
|
搜索推荐 算法 Java
|
12天前
|
机器学习/深度学习 运维 算法
深入探索机器学习中的支持向量机(SVM)算法:原理、应用与Python代码示例全面解析
【8月更文挑战第6天】在机器学习领域,支持向量机(SVM)犹如璀璨明珠。它是一种强大的监督学习算法,在分类、回归及异常检测中表现出色。SVM通过在高维空间寻找最大间隔超平面来分隔不同类别的数据,提升模型泛化能力。为处理非线性问题,引入了核函数将数据映射到高维空间。SVM在文本分类、图像识别等多个领域有广泛应用,展现出高度灵活性和适应性。
65 2
|
16天前
|
算法 5G vr&ar
基于1bitDAC的MU-MIMO的非线性预编码算法matlab性能仿真
在现代无线通信中,1-bit DAC的非线性预编码技术应用于MU-MIMO系统,旨在降低成本与能耗。本文采用MATLAB 2022a版本,深入探讨此技术,并通过算法运行效果图展示性能。核心代码支持中文注释与操作指导。理论部分包括信号量化、符号最大化准则,并对比ZF、WF、MRT及ADMM等算法,揭示了在1-bit量化条件下如何优化预编码以提升系统性能。
|
26天前
|
机器学习/深度学习 存储 算法
编码之舞:从算法到应用的探索之旅
在数字化时代的浪潮中,编程技术如同一种语言,连接着人类与机器。本文将带领读者踏上一场自数据结构基础至高级算法应用的探索旅程,通过实际案例分析,揭示算法在现代软件开发中的重要作用,并分享作者在编程实践中的心得体会,旨在为初学者和资深开发者提供有价值的参考与启示。
|
5天前
|
搜索推荐 算法 Java
插入排序算法(Java代码实现)
这篇文章通过Java代码示例详细解释了插入排序算法的实现过程,包括算法的基本思想、核心代码、辅助函数以及测试结果,展示了如何通过插入排序对数组进行升序排列。
|
7天前
|
机器学习/深度学习 算法 Python
python与朴素贝叶斯算法(附示例和代码)
朴素贝叶斯算法以其高效性和优良的分类性能,成为文本处理领域一项受欢迎的方法。提供的代码示例证明了其在Python语言中的易用性和实用性。尽管算法假设了特征之间的独立性,但在实际应用中,它仍然能够提供强大的分类能力。通过调整参数和优化模型,你可以进一步提升朴素贝叶斯分类器的性能。
17 0
|
14天前
|
算法 C++
惊爆!KPM算法背后的秘密武器:一行代码揭秘字符串最小周期的终极奥义,让你秒变编程界周期大师!
【8月更文挑战第4天】字符串最小周期问题旨在找出字符串中最短重复子串的长度。KPM(实为KMP,Knuth-Morris-Pratt)算法,虽主要用于字符串匹配,但其生成的前缀函数(next数组)也可用于求解最小周期。核心思想是构建LPS数组,记录模式串中每个位置的最长相等前后缀长度。对于长度为n的字符串S,其最小周期T可通过公式ans = n - LPS[n-1]求得。通过分析周期字符串的特性,可证明该方法的有效性。提供的C++示例代码展示了如何计算给定字符串的最小周期,体现了KPM算法在解决此类问题上的高效性。
25 0
|
28天前
|
算法 Java 调度
高并发架构设计三大利器:缓存、限流和降级问题之使用Java代码实现令牌桶算法问题如何解决
高并发架构设计三大利器:缓存、限流和降级问题之使用Java代码实现令牌桶算法问题如何解决

热门文章

最新文章