📄 mpegaudio.c
字号:
printf("%2d:%d in=%x %x %d\n",
j, i, vmax, scale_factor_table[index], index);
#endif
/* store the scale factor */
assert(index >=0 && index <= 63);
sf[i] = index;
}
/* compute the transmission factor : look if the scale factors
are close enough to each other */
d1 = scale_diff_table[sf[0] - sf[1] + 64];
d2 = scale_diff_table[sf[1] - sf[2] + 64];
/* handle the 25 cases */
switch(d1 * 5 + d2) {
case 0*5+0:
case 0*5+4:
case 3*5+4:
case 4*5+0:
case 4*5+4:
code = 0;
break;
case 0*5+1:
case 0*5+2:
case 4*5+1:
case 4*5+2:
code = 3;
sf[2] = sf[1];
break;
case 0*5+3:
case 4*5+3:
code = 3;
sf[1] = sf[2];
break;
case 1*5+0:
case 1*5+4:
case 2*5+4:
code = 1;
sf[1] = sf[0];
break;
case 1*5+1:
case 1*5+2:
case 2*5+0:
case 2*5+1:
case 2*5+2:
code = 2;
sf[1] = sf[2] = sf[0];
break;
case 2*5+3:
case 3*5+3:
code = 2;
sf[0] = sf[1] = sf[2];
break;
case 3*5+0:
case 3*5+1:
case 3*5+2:
code = 2;
sf[0] = sf[2] = sf[1];
break;
case 1*5+3:
code = 2;
if (sf[0] > sf[2])
sf[0] = sf[2];
sf[1] = sf[2] = sf[0];
break;
default:
abort();
}
#if 0
printf("%d: %2d %2d %2d %d %d -> %d\n", j,
sf[0], sf[1], sf[2], d1, d2, code);
#endif
scale_code[j] = code;
sf += 3;
}
}
/* The most important function : psycho acoustic module. In this
encoder there is basically none, so this is the worst you can do,
but also this is the simpler. */
static void psycho_acoustic_model(MpegAudioContext *s, short smr[SBLIMIT])
{
int i;
for(i=0;i<s->sblimit;i++) {
smr[i] = (int)(fixed_smr[i] * 10);
}
}
#define SB_NOTALLOCATED 0
#define SB_ALLOCATED 1
#define SB_NOMORE 2
/* Try to maximize the smr while using a number of bits inferior to
the frame size. I tried to make the code simpler, faster and
smaller than other encoders :-) */
static void compute_bit_allocation(MpegAudioContext *s,
short smr1[MPA_MAX_CHANNELS][SBLIMIT],
unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
int *padding)
{
int i, ch, b, max_smr, max_ch, max_sb, current_frame_size, max_frame_size;
int incr;
short smr[MPA_MAX_CHANNELS][SBLIMIT];
unsigned char subband_status[MPA_MAX_CHANNELS][SBLIMIT];
const unsigned char *alloc;
memcpy(smr, smr1, s->nb_channels * sizeof(short) * SBLIMIT);
memset(subband_status, SB_NOTALLOCATED, s->nb_channels * SBLIMIT);
memset(bit_alloc, 0, s->nb_channels * SBLIMIT);
/* compute frame size and padding */
max_frame_size = s->frame_size;
s->frame_frac += s->frame_frac_incr;
if (s->frame_frac >= 65536) {
s->frame_frac -= 65536;
s->do_padding = 1;
max_frame_size += 8;
} else {
s->do_padding = 0;
}
/* compute the header + bit alloc size */
current_frame_size = 32;
alloc = s->alloc_table;
for(i=0;i<s->sblimit;i++) {
incr = alloc[0];
current_frame_size += incr * s->nb_channels;
alloc += 1 << incr;
}
for(;;) {
/* look for the subband with the largest signal to mask ratio */
max_sb = -1;
max_ch = -1;
max_smr = 0x80000000;
for(ch=0;ch<s->nb_channels;ch++) {
for(i=0;i<s->sblimit;i++) {
if (smr[ch][i] > max_smr && subband_status[ch][i] != SB_NOMORE) {
max_smr = smr[ch][i];
max_sb = i;
max_ch = ch;
}
}
}
#if 0
printf("current=%d max=%d max_sb=%d alloc=%d\n",
current_frame_size, max_frame_size, max_sb,
bit_alloc[max_sb]);
#endif
if (max_sb < 0)
break;
/* find alloc table entry (XXX: not optimal, should use
pointer table) */
alloc = s->alloc_table;
for(i=0;i<max_sb;i++) {
alloc += 1 << alloc[0];
}
if (subband_status[max_ch][max_sb] == SB_NOTALLOCATED) {
/* nothing was coded for this band: add the necessary bits */
incr = 2 + nb_scale_factors[s->scale_code[max_ch][max_sb]] * 6;
incr += total_quant_bits[alloc[1]];
} else {
/* increments bit allocation */
b = bit_alloc[max_ch][max_sb];
incr = total_quant_bits[alloc[b + 1]] -
total_quant_bits[alloc[b]];
}
if (current_frame_size + incr <= max_frame_size) {
/* can increase size */
b = ++bit_alloc[max_ch][max_sb];
current_frame_size += incr;
/* decrease smr by the resolution we added */
smr[max_ch][max_sb] = smr1[max_ch][max_sb] - quant_snr[alloc[b]];
/* max allocation size reached ? */
if (b == ((1 << alloc[0]) - 1))
subband_status[max_ch][max_sb] = SB_NOMORE;
else
subband_status[max_ch][max_sb] = SB_ALLOCATED;
} else {
/* cannot increase the size of this subband */
subband_status[max_ch][max_sb] = SB_NOMORE;
}
}
*padding = max_frame_size - current_frame_size;
assert(*padding >= 0);
#if 0
for(i=0;i<s->sblimit;i++) {
printf("%d ", bit_alloc[i]);
}
printf("\n");
#endif
}
/*
* Output the mpeg audio layer 2 frame. Note how the code is small
* compared to other encoders :-)
*/
static void encode_frame(MpegAudioContext *s,
unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
int padding)
{
int i, j, k, l, bit_alloc_bits, b, ch;
unsigned char *sf;
int q[3];
PutBitContext *p = &s->pb;
/* header */
put_bits(p, 12, 0xfff);
put_bits(p, 1, 1 - s->lsf); /* 1 = mpeg1 ID, 0 = mpeg2 lsf ID */
put_bits(p, 2, 4-2); /* layer 2 */
put_bits(p, 1, 1); /* no error protection */
put_bits(p, 4, s->bitrate_index);
put_bits(p, 2, s->freq_index);
put_bits(p, 1, s->do_padding); /* use padding */
put_bits(p, 1, 0); /* private_bit */
put_bits(p, 2, s->nb_channels == 2 ? MPA_STEREO : MPA_MONO);
put_bits(p, 2, 0); /* mode_ext */
put_bits(p, 1, 0); /* no copyright */
put_bits(p, 1, 1); /* original */
put_bits(p, 2, 0); /* no emphasis */
/* bit allocation */
j = 0;
for(i=0;i<s->sblimit;i++) {
bit_alloc_bits = s->alloc_table[j];
for(ch=0;ch<s->nb_channels;ch++) {
put_bits(p, bit_alloc_bits, bit_alloc[ch][i]);
}
j += 1 << bit_alloc_bits;
}
/* scale codes */
for(i=0;i<s->sblimit;i++) {
for(ch=0;ch<s->nb_channels;ch++) {
if (bit_alloc[ch][i])
put_bits(p, 2, s->scale_code[ch][i]);
}
}
/* scale factors */
for(i=0;i<s->sblimit;i++) {
for(ch=0;ch<s->nb_channels;ch++) {
if (bit_alloc[ch][i]) {
sf = &s->scale_factors[ch][i][0];
switch(s->scale_code[ch][i]) {
case 0:
put_bits(p, 6, sf[0]);
put_bits(p, 6, sf[1]);
put_bits(p, 6, sf[2]);
break;
case 3:
case 1:
put_bits(p, 6, sf[0]);
put_bits(p, 6, sf[2]);
break;
case 2:
put_bits(p, 6, sf[0]);
break;
}
}
}
}
/* quantization & write sub band samples */
for(k=0;k<3;k++) {
for(l=0;l<12;l+=3) {
j = 0;
for(i=0;i<s->sblimit;i++) {
bit_alloc_bits = s->alloc_table[j];
for(ch=0;ch<s->nb_channels;ch++) {
b = bit_alloc[ch][i];
if (b) {
int qindex, steps, m, sample, bits;
/* we encode 3 sub band samples of the same sub band at a time */
qindex = s->alloc_table[j+b];
steps = quant_steps[qindex];
for(m=0;m<3;m++) {
sample = s->sb_samples[ch][k][l + m][i];
/* divide by scale factor */
#ifdef USE_FLOATS
{
float a;
a = (float)sample * scale_factor_inv_table[s->scale_factors[ch][i][k]];
q[m] = (int)((a + 1.0) * steps * 0.5);
}
#else
{
int q1, e, shift, mult;
e = s->scale_factors[ch][i][k];
shift = scale_factor_shift[e];
mult = scale_factor_mult[e];
/* normalize to P bits */
if (shift < 0)
q1 = sample << (-shift);
else
q1 = sample >> shift;
q1 = (q1 * mult) >> P;
q[m] = ((q1 + (1 << P)) * steps) >> (P + 1);
}
#endif
if (q[m] >= steps)
q[m] = steps - 1;
assert(q[m] >= 0 && q[m] < steps);
}
bits = quant_bits[qindex];
if (bits < 0) {
/* group the 3 values to save bits */
put_bits(p, -bits,
q[0] + steps * (q[1] + steps * q[2]));
#if 0
printf("%d: gr1 %d\n",
i, q[0] + steps * (q[1] + steps * q[2]));
#endif
} else {
#if 0
printf("%d: gr3 %d %d %d\n",
i, q[0], q[1], q[2]);
#endif
put_bits(p, bits, q[0]);
put_bits(p, bits, q[1]);
put_bits(p, bits, q[2]);
}
}
}
/* next subband in alloc table */
j += 1 << bit_alloc_bits;
}
}
}
/* padding */
for(i=0;i<padding;i++)
put_bits(p, 1, 0);
/* flush */
flush_put_bits(p);
}
int MPA_encode_frame(AVEncodeContext *avctx,
unsigned char *frame, int buf_size, void *data)
{
MpegAudioContext *s = avctx->priv_data;
short *samples = data;
short smr[MPA_MAX_CHANNELS][SBLIMIT];
unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT];
int padding, i;
for(i=0;i<s->nb_channels;i++) {
filter(s, i, samples + i, s->nb_channels);
}
for(i=0;i<s->nb_channels;i++) {
compute_scale_factors(s->scale_code[i], s->scale_factors[i],
s->sb_samples[i], s->sblimit);
}
for(i=0;i<s->nb_channels;i++) {
psycho_acoustic_model(s, smr[i]);
}
compute_bit_allocation(s, smr, bit_alloc, &padding);
init_put_bits(&s->pb, frame, MPA_MAX_CODED_FRAME_SIZE, NULL, NULL);
encode_frame(s, bit_alloc, padding);
s->nb_samples += MPA_FRAME_SIZE;
return s->pb.buf_ptr - s->pb.buf;
}
AVEncoder mp2_encoder = {
"mp2",
CODEC_TYPE_AUDIO,
CODEC_ID_MP2,
sizeof(MpegAudioContext),
MPA_encode_init,
MPA_encode_frame,
NULL,
};
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -