view libfaad2/ssr_fb.c @ 23510:a6c619ee9d30

Teletext support for tv:// (v4l and v4l2 only) modified patch from Otvos Attila oattila at chello dot hu Module uses zvbi library for all low-level VBI operations (like I/O with vbi device, converting vbi pages into usefull vbi_page stuctures, rendering them into RGB32 images). All teletext related stuff (except properties, slave commands and rendering osd in text mode or RGB32 rendered teletext pages in spu mode) is implemented in tvi_vbi.c New properties: teletext_page - switching between pages teletext_mode - switch between on/off/opaque/transparent modes teletext_format - (currently read-only) allows to get format info (black/white,gray,text) teletext_half_page - trivial zooming (displaying top/bottom half of teletext page) New slave commands: teletext_add_dec - user interface for jumping to any page by editing page number interactively teletext_go_link - goes though links, specified on current page
author voroshil
date Sun, 10 Jun 2007 00:06:12 +0000
parents 59b6fa5b4201
children e83eef58b30a
line wrap: on
line source

/*
** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
** Copyright (C) 2003-2004 M. Bakker, Ahead Software AG, http://www.nero.com
**  
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
** 
** This program is distributed in the hope that it will be useful,
** but WITHOUT ANY WARRANTY; without even the implied warranty of
** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
** GNU General Public License for more details.
** 
** You should have received a copy of the GNU General Public License
** along with this program; if not, write to the Free Software 
** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
**
** Any non-GPL usage of this software or parts of this software is strictly
** forbidden.
**
** Commercial non-GPL licensing of this software is possible.
** For more info contact Ahead Software through Mpeg4AAClicense@nero.com.
**
** $Id: ssr_fb.c,v 1.13 2004/09/04 14:56:29 menno Exp $
**/

#include "common.h"
#include "structs.h"

#ifdef SSR_DEC

#include <string.h>
#include <stdlib.h>
#include "syntax.h"
#include "filtbank.h"
#include "mdct.h"
#include "ssr_fb.h"
#include "ssr_win.h"

fb_info *ssr_filter_bank_init(uint16_t frame_len)
{
    uint16_t nshort = frame_len/8;

    fb_info *fb = (fb_info*)faad_malloc(sizeof(fb_info));
    memset(fb, 0, sizeof(fb_info));

    /* normal */
    fb->mdct256 = faad_mdct_init(2*nshort);
    fb->mdct2048 = faad_mdct_init(2*frame_len);

    fb->long_window[0]  = sine_long_256;
    fb->short_window[0] = sine_short_32;
    fb->long_window[1]  = kbd_long_256;
    fb->short_window[1] = kbd_short_32;

    return fb;
}

void ssr_filter_bank_end(fb_info *fb)
{
    faad_mdct_end(fb->mdct256);
    faad_mdct_end(fb->mdct2048);

    if (fb) faad_free(fb);
}

static INLINE void imdct_ssr(fb_info *fb, real_t *in_data,
                             real_t *out_data, uint16_t len)
{
    mdct_info *mdct;

    switch (len)
    {
    case 512:
        mdct = fb->mdct2048;
        break;
    case 64:
        mdct = fb->mdct256;
        break;
    }

    faad_imdct(mdct, in_data, out_data);
}

/* NON-overlapping inverse filterbank for use with SSR */
void ssr_ifilter_bank(fb_info *fb, uint8_t window_sequence, uint8_t window_shape,
                      uint8_t window_shape_prev, real_t *freq_in,
                      real_t *time_out, uint16_t frame_len)
{
    int16_t i;
    real_t *transf_buf;

    real_t *window_long;
    real_t *window_long_prev;
    real_t *window_short;
    real_t *window_short_prev;

    uint16_t nlong = frame_len;
    uint16_t nshort = frame_len/8;
    uint16_t trans = nshort/2;

    uint16_t nflat_ls = (nlong-nshort)/2;

    transf_buf = (real_t*)faad_malloc(2*nlong*sizeof(real_t));

    window_long       = fb->long_window[window_shape];
    window_long_prev  = fb->long_window[window_shape_prev];
    window_short      = fb->short_window[window_shape];
    window_short_prev = fb->short_window[window_shape_prev];

    switch (window_sequence)
    {
    case ONLY_LONG_SEQUENCE:
        imdct_ssr(fb, freq_in, transf_buf, 2*nlong);
        for (i = nlong-1; i >= 0; i--)
        {
            time_out[i] = MUL_R_C(transf_buf[i],window_long_prev[i]);
            time_out[nlong+i] = MUL_R_C(transf_buf[nlong+i],window_long[nlong-1-i]);
        }
        break;

    case LONG_START_SEQUENCE:
        imdct_ssr(fb, freq_in, transf_buf, 2*nlong);
        for (i = 0; i < nlong; i++)
            time_out[i] = MUL_R_C(transf_buf[i],window_long_prev[i]);
        for (i = 0; i < nflat_ls; i++)
            time_out[nlong+i] = transf_buf[nlong+i];
        for (i = 0; i < nshort; i++)
            time_out[nlong+nflat_ls+i] = MUL_R_C(transf_buf[nlong+nflat_ls+i],window_short[nshort-i-1]);
        for (i = 0; i < nflat_ls; i++)
            time_out[nlong+nflat_ls+nshort+i] = 0;
        break;

    case EIGHT_SHORT_SEQUENCE:
        imdct_ssr(fb, freq_in+0*nshort, transf_buf+2*nshort*0, 2*nshort);
        imdct_ssr(fb, freq_in+1*nshort, transf_buf+2*nshort*1, 2*nshort);
        imdct_ssr(fb, freq_in+2*nshort, transf_buf+2*nshort*2, 2*nshort);
        imdct_ssr(fb, freq_in+3*nshort, transf_buf+2*nshort*3, 2*nshort);
        imdct_ssr(fb, freq_in+4*nshort, transf_buf+2*nshort*4, 2*nshort);
        imdct_ssr(fb, freq_in+5*nshort, transf_buf+2*nshort*5, 2*nshort);
        imdct_ssr(fb, freq_in+6*nshort, transf_buf+2*nshort*6, 2*nshort);
        imdct_ssr(fb, freq_in+7*nshort, transf_buf+2*nshort*7, 2*nshort);
        for(i = nshort-1; i >= 0; i--)
        {
            time_out[i+0*nshort] = MUL_R_C(transf_buf[nshort*0+i],window_short_prev[i]);
            time_out[i+1*nshort] = MUL_R_C(transf_buf[nshort*1+i],window_short[i]);
            time_out[i+2*nshort] = MUL_R_C(transf_buf[nshort*2+i],window_short_prev[i]);
            time_out[i+3*nshort] = MUL_R_C(transf_buf[nshort*3+i],window_short[i]);
            time_out[i+4*nshort] = MUL_R_C(transf_buf[nshort*4+i],window_short_prev[i]);
            time_out[i+5*nshort] = MUL_R_C(transf_buf[nshort*5+i],window_short[i]);
            time_out[i+6*nshort] = MUL_R_C(transf_buf[nshort*6+i],window_short_prev[i]);
            time_out[i+7*nshort] = MUL_R_C(transf_buf[nshort*7+i],window_short[i]);
            time_out[i+8*nshort] = MUL_R_C(transf_buf[nshort*8+i],window_short_prev[i]);
            time_out[i+9*nshort] = MUL_R_C(transf_buf[nshort*9+i],window_short[i]);
            time_out[i+10*nshort] = MUL_R_C(transf_buf[nshort*10+i],window_short_prev[i]);
            time_out[i+11*nshort] = MUL_R_C(transf_buf[nshort*11+i],window_short[i]);
            time_out[i+12*nshort] = MUL_R_C(transf_buf[nshort*12+i],window_short_prev[i]);
            time_out[i+13*nshort] = MUL_R_C(transf_buf[nshort*13+i],window_short[i]);
            time_out[i+14*nshort] = MUL_R_C(transf_buf[nshort*14+i],window_short_prev[i]);
            time_out[i+15*nshort] = MUL_R_C(transf_buf[nshort*15+i],window_short[i]);
        }
        break;

    case LONG_STOP_SEQUENCE:
        imdct_ssr(fb, freq_in, transf_buf, 2*nlong);
        for (i = 0; i < nflat_ls; i++)
            time_out[i] = 0;
        for (i = 0; i < nshort; i++)
            time_out[nflat_ls+i] = MUL_R_C(transf_buf[nflat_ls+i],window_short_prev[i]);
        for (i = 0; i < nflat_ls; i++)
            time_out[nflat_ls+nshort+i] = transf_buf[nflat_ls+nshort+i];
        for (i = 0; i < nlong; i++)
            time_out[nlong+i] = MUL_R_C(transf_buf[nlong+i],window_long[nlong-1-i]);
		break;
    }

    faad_free(transf_buf);
}


#endif