mp4evops.c

Intel提供的在PCA(例如Sitsang开发板)使用IPP4.0库的MPEG-4编码源程序

/******************************************************************************
//               INTEL CORPORATION PROPRIETARY INFORMATION
//  This software is supplied under the terms of a license agreement or
//  nondisclosure agreement with Intel Corporation and may not be copied
//  or disclosed except in accordance with the terms of that agreement.
//        Copyright (c) 2003 Intel Corporation. All Rights Reserved.
//
//  Description:    VOP or frame encoding functions of MPEG-4 video encoder
//                  sample code for Intel(R) Integrated Performance Primitives.
//  Functions List:
//		encode_pvop_mpeg4()
//		encode_ivop_mpeg4()
//		encode_mpeg4()
******************************************************************************/

#include "sampmp4.h"

/******************************************************************************
// Name:		encode_pvop_mpeg4
// Desicription:
//      Encode the PVOP data in non-ER mode
//
//	Input Arguments:
//      stream_buf  - Pointer to the output video bitstream
//      enc_state   - Pointer to the general state struct of MPEG-4 encoder
//      vop_infor   - Pointer to the vop information struct of MPEG-4 encoder
//
//	Output Arguments:
//      stream_buf  - Pointer to the updated output video bitstream after
//                    compression
//      enc_state   - Pointer to the updated general state struct of MPEG-4
//                    encoder
//      vop_infor   - Pointer to the updated vop information struct of MPEG-4
//                    encoder
//
//	Returns:
//      SAMPLE_STATUS_NOERR      If succeeds
//      SAMPLE_STATUS_ERR        If encoding fails
//  Note:
//      Motion estimation based on macroblocks are performed in this function.
******************************************************************************/
sample_status encode_pvop_mpeg4(sample_bitstream  *stream_buf,
                                mp4_enc_state     *enc_state,
                                mp4_enc_vop_infor *vop_infor)
{
	Ipp16u      *src_sum_mb;
	Ipp8u       *y_plane;
    Ipp8u		*src_cur_mb, *src_ref_mb, *src_ref_rec_mb, *src_ref_ori_mb;

    Ipp8u		transp_curmb_buf[4 + 4];
    Ipp8u*      transp_curmb = (Ipp8u*)SAMPLE_ALIGN4(transp_curmb_buf);
	Ipp8u		tmp_cur_mb_buf[64 * 6 + 8];
	Ipp8u*      tmp_cur_mb = (Ipp8u*)SAMPLE_ALIGN8(tmp_cur_mb_buf);    
    Ipp8u*      type_cur_mb;

    int			mbx_indx, mby_indx, i, j, y_step;
	int			mad = 0;
    IppiRect    ref_rect;

    IppMotionVector *mv_cur_mb, mv_pred;
    IppMotionVector *mv_left_mb, *mv_top_mb, *mv_top_right_mb;
	sample_status    ret_code;
    
	/* initialize high-level encoder state */
	ret_code = init_vop_infor_enc_mpeg4(enc_state, vop_infor);
	if (SAMPLE_STATUS_NOERR != ret_code ) {
		return ret_code;
	}
    
    /* for rectangular shape, all four blocks in one macroblock are not
    // transparent */
	for(i = 0; i < 4; i++) {
		transp_curmb[i]   = IPP_VIDEO_OPAQUE;
	}
    
	ref_rect.x	= - SAMPLE_VIDEO_MB_SIZE;
	ref_rect.y	= - SAMPLE_VIDEO_MB_SIZE;
	ref_rect.height	= enc_state->frame_dimension.height + 2 * SAMPLE_VIDEO_MB_SIZE;
	ref_rect.width	= enc_state->frame_dimension.width  + 2 * SAMPLE_VIDEO_MB_SIZE;

    /* forward reference plane */
    y_step  = enc_state->frame_step_set.y_step;
	y_plane = enc_state->fwd_ref_frame.y_ptr - 16 - 16 * y_step;

    /* compute summation plane based on blocks of 4 * 4 pixels in forward
    // reference plane */
    /* y_plane and enc_state->ysum_plane must be 8 bytes aligned */
    ret_code = ippiSumNorm_VOP_MPEG4_8u16u(
        y_plane,
        &ref_rect,
        enc_state->ysum_plane,
        5,
        y_step);
    
    if (ippStsNoErr != ret_code) {
        return SAMPLE_STATUS_ERR;
    }

    mv_cur_mb       = enc_state->mv_plane;
    type_cur_mb     = enc_state->mb_type_plane;
    src_sum_mb      = enc_state->ysum_plane + 16 + 16 * y_step;
    src_cur_mb      = enc_state->cur_frame.y_ptr;
    src_ref_ori_mb  = enc_state->fwd_ref_frame.y_ptr;
    src_ref_rec_mb  = enc_state->fwd_ref_rec_frame.y_ptr;
    src_ref_mb      = (enc_state->use_src_me == 1) ? src_ref_ori_mb : NULL;
    
	for(mby_indx = 0; mby_indx < enc_state->mb_per_col; mby_indx++) {

		for(mbx_indx = 0; mbx_indx < enc_state->mb_per_row; mbx_indx++) {

            /* locate three neighbourhood macroblocks for candidate predictors
            // Please refer to subclause 7.6.5 of ISO/IEC 14496-2:2001(E) */
			mv_left_mb      = mv_cur_mb	- 4;
			mv_top_mb       = mv_cur_mb	- 4 * (enc_state->mb_per_row + 2);
			mv_top_right_mb	= mv_top_mb	+ 4;

            /* form the motion vector predictor by a median filtering of three
            // candidate predictors for the first block in one macroblock */
            /* mv_cur_mb, mv_left_mb, mv_top_mb, mv_top_right_mb must be
            // 4 bytes aligned */
			ret_code = ippiFindMVpred_MPEG4(
                mv_cur_mb,
                mv_left_mb,
                mv_top_mb,
                mv_top_right_mb,
                vop_infor->tranp_buf,
                vop_infor->tranp_buf + 4,
                vop_infor->tranp_buf + 8,
                transp_curmb,
                &mv_pred,
                NULL,
                0);
            if (ippStsNoErr != ret_code) {
                return ret_code;
            }

            /* copy the macroblock in the plane to a contiguous 16*16 buffer */
			for(i = 0; i < 16; i++) {
                for (j = 0; j < 16; j++) {
                    tmp_cur_mb[i * 16 + j] = src_cur_mb[y_step * i + j];
                }
			}
			
            /* perform 16x16 size block match with Successive Elimination
            // Algorithm */
            /* src_ref_mb, src_ref_rec_mb, src_sum_mb, tmp_cur_mb must be
            // 8 bytes aligned */            
            ret_code = ippiMotionEstimation_16x16_SEA(
                src_ref_mb,
                src_ref_rec_mb, 				
                src_sum_mb,
                tmp_cur_mb,
                &ref_rect,
                &vop_infor->cur_point,
                &mv_pred,
                mv_cur_mb,
                type_cur_mb,
                &mad,
                y_step,
                enc_state->rounding,
                enc_state->search_range,
                5);
            
            if (ippStsNoErr != ret_code) {
                return SAMPLE_STATUS_ERR;
            }
            
            /* update pointers for next macroblock in the same row */
            src_sum_mb	    += SAMPLE_VIDEO_MB_SIZE;
			src_cur_mb      += SAMPLE_VIDEO_MB_SIZE;
            src_ref_rec_mb  += SAMPLE_VIDEO_MB_SIZE;            
            if (src_ref_mb) {
                src_ref_mb  += SAMPLE_VIDEO_MB_SIZE;
            }

            vop_infor->tranp_buf += 4;
			if (0 == mby_indx) {
				for(i=0;i<4;i++){
					vop_infor->tranp_buf[i] = IPP_VIDEO_OPAQUE;
				}
			}

			vop_infor->cur_point.x += 16;
			vop_infor->ref_point.x += 16;

            mv_cur_mb += 4;
            type_cur_mb ++;

        } /* loop for each low */
        
        /* update pointers for first macroblock in the next row */
        src_sum_mb     += 16 * y_step - 16 * enc_state->mb_per_row;
        src_cur_mb     += 16 * y_step - 16 * enc_state->mb_per_row;
        src_ref_rec_mb += 16 * y_step - 16 * enc_state->mb_per_row;
        if (src_ref_mb) {
            src_ref_mb += 16 * y_step - 16 * enc_state->mb_per_row;
        }
        
        vop_infor->tranp_buf = enc_state->tranp_buf;

		vop_infor->cur_point.x = 0;
		vop_infor->ref_point.x = 0;
		vop_infor->cur_point.y += 16;
		vop_infor->ref_point.y += 16;

		mv_cur_mb += 2 * 4;        
    } /* loop for each column */

    /* dump vop header information into output stream */
	ret_code = create_vop_header_mpeg4 (stream_buf, enc_state, vop_infor);
	if (SAMPLE_STATUS_NOERR != ret_code) {
		return ret_code;
	}

	/* begin texture encoding */
	for(i=0; i< (enc_state->mb_per_row + 2) * 4; i++ ) {
		vop_infor->tranp_buf[i]   = IPP_VIDEO_TRANSPARENT;
	}
    
    vop_infor->cur_point.x = 0;
	vop_infor->ref_point.x = 0;
	vop_infor->cur_point.y = 0;
	vop_infor->ref_point.y = 0;

    mv_cur_mb   = enc_state->mv_plane;
    type_cur_mb = enc_state->mb_type_plane;

    /**************************************************************************
    // begin the encoding of PVOP, MB by MB. notes the pointers should be set
    // or reset at the beginning and ending per Row
    **************************************************************************/
	for(mby_indx = 0; mby_indx < enc_state -> mb_per_col; mby_indx ++){

		for(mbx_indx = 0; mbx_indx < enc_state -> mb_per_row ; mbx_indx ++) {

			vop_infor->mb_type = *type_cur_mb;
            
            if ((IPP_VIDEO_INTRA == vop_infor->mb_type)
                || (IPP_VIDEO_INTRA_Q == vop_infor->mb_type)) {
                ret_code = encode_intra_mb_mpeg4(stream_buf, enc_state, vop_infor);
                if (SAMPLE_STATUS_NOERR != ret_code) {
                    return ret_code;
                }
            } else {
                ret_code = encode_inter_mb_mpeg4(stream_buf, enc_state, vop_infor, mv_cur_mb, transp_curmb);
                if (SAMPLE_STATUS_NOERR != ret_code) {
                    return ret_code;
                }
            }
            /* update for the next macroblck in the same row */
            vop_infor->mb_indx ++;

            vop_infor->cur_mb.y_ptr  += SAMPLE_VIDEO_MB_SIZE;
			vop_infor->cur_mb.cb_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
			vop_infor->cur_mb.cr_ptr += SAMPLE_VIDEO_BLOCK_SIZE;

			vop_infor->fwd_ref_mb.y_ptr  += SAMPLE_VIDEO_MB_SIZE;
			vop_infor->fwd_ref_mb.cb_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
			vop_infor->fwd_ref_mb.cr_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
					   			
			vop_infor->rec_mb.y_ptr  += SAMPLE_VIDEO_MB_SIZE;
			vop_infor->rec_mb.cb_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
			vop_infor->rec_mb.cr_ptr += SAMPLE_VIDEO_BLOCK_SIZE;

			vop_infor->fwd_ref_rec_mb.y_ptr  += SAMPLE_VIDEO_MB_SIZE;
			vop_infor->fwd_ref_rec_mb.cb_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
			vop_infor->fwd_ref_rec_mb.cr_ptr += SAMPLE_VIDEO_BLOCK_SIZE;

            vop_infor->coef_buf_row.y_ptr  += SAMPLE_VIDEO_MB_SIZE;
            vop_infor->coef_buf_row.cb_ptr += SAMPLE_VIDEO_BLOCK_SIZE;
            vop_infor->coef_buf_row.cr_ptr += SAMPLE_VIDEO_BLOCK_SIZE;