54166165 |
/*
* SVQ1 Encoder
* Copyright (C) 2004 Mike Melanson <melanson@pcisys.net>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/** |
ba87f080 |
* @file |
54166165 |
* Sorenson Vector Quantizer #1 (SVQ1) video codec.
* For more information of the SVQ1 algorithm, visit:
* http://www.pcisys.net/~melanson/codecs/
*/
#include "avcodec.h" |
b42d594c |
#include "hpeldsp.h" |
2d604443 |
#include "me_cmp.h" |
54166165 |
#include "mpegvideo.h" |
fc53b6af |
#include "h263.h" |
7ffd8332 |
#include "internal.h" |
e0c16e4e |
#include "mpegutils.h" |
54166165 |
#include "svq1.h" |
65d5d586 |
#include "svq1enc.h" |
54166165 |
#include "svq1enc_cb.h" |
9517900b |
#include "libavutil/avassert.h" |
54166165 |
|
2ea2612d |
static void svq1_write_header(SVQ1EncContext *s, int frame_type) |
54166165 |
{
int i;
/* frame code */
put_bits(&s->pb, 22, 0x20);
/* temporal reference (sure hope this is a "don't care") */
put_bits(&s->pb, 8, 0x00);
/* frame type */
put_bits(&s->pb, 2, frame_type - 1);
|
ce5e49b0 |
if (frame_type == AV_PICTURE_TYPE_I) { |
54166165 |
/* no checksum since frame code is 0x20 */
/* no embedded string either */
/* output 5 unknown bits (2 + 2 + 1) */
put_bits(&s->pb, 5, 2); /* 2 needed by quicktime decoder */
|
de31814a |
i = ff_match_2uint16((void*)ff_svq1_frame_size_table, |
ada12f83 |
FF_ARRAY_ELEMS(ff_svq1_frame_size_table),
s->frame_width, s->frame_height); |
c46eeae2 |
put_bits(&s->pb, 3, i); |
54166165 |
|
ada12f83 |
if (i == 7) {
put_bits(&s->pb, 12, s->frame_width);
put_bits(&s->pb, 12, s->frame_height); |
54166165 |
}
}
/* no checksum or extra data (next 2 bits get 0) */
put_bits(&s->pb, 2, 0);
}
|
ada12f83 |
#define QUALITY_THRESHOLD 100 |
54166165 |
#define THRESHOLD_MULTIPLIER 0.6
|
65d5d586 |
static int ssd_int8_vs_int16_c(const int8_t *pix1, const int16_t *pix2, |
dad31083 |
intptr_t size) |
65d5d586 |
{
int score = 0, i;
for (i = 0; i < size; i++)
score += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]);
return score;
}
|
2ea2612d |
static int encode_block(SVQ1EncContext *s, uint8_t *src, uint8_t *ref, |
ada12f83 |
uint8_t *decoded, int stride, int level,
int threshold, int lambda, int intra)
{ |
54166165 |
int count, y, x, i, j, split, best_mean, best_score, best_count;
int best_vector[6]; |
ada12f83 |
int block_sum[7] = { 0, 0, 0, 0, 0, 0 }; |
63a46c61 |
int w = 2 << (level + 2 >> 1);
int h = 2 << (level + 1 >> 1); |
ada12f83 |
int size = w * h; |
d0682b5e |
int16_t (*block)[256] = s->encoded_block_levels[level]; |
54166165 |
const int8_t *codebook_sum, *codebook; |
ada12f83 |
const uint16_t(*mean_vlc)[2];
const uint8_t(*multistage_vlc)[2];
best_score = 0;
// FIXME: Optimize, this does not need to be done multiple times.
if (intra) { |
9079e99d |
// level is 5 when encode_block is called from svq1_encode_plane
// and always < 4 when called recursively from this function.
codebook_sum = level < 4 ? svq1_intra_codebook_sum[level] : NULL; |
ada12f83 |
codebook = ff_svq1_intra_codebooks[level];
mean_vlc = ff_svq1_intra_mean_vlc;
multistage_vlc = ff_svq1_intra_multistage_vlc[level];
for (y = 0; y < h; y++) {
for (x = 0; x < w; x++) {
int v = src[x + y * stride];
block[0][x + w * y] = v;
best_score += v * v;
block_sum[0] += v; |
54166165 |
}
} |
ada12f83 |
} else { |
9079e99d |
// level is 5 or < 4, see above for details.
codebook_sum = level < 4 ? svq1_inter_codebook_sum[level] : NULL; |
ada12f83 |
codebook = ff_svq1_inter_codebooks[level];
mean_vlc = ff_svq1_inter_mean_vlc + 256;
multistage_vlc = ff_svq1_inter_multistage_vlc[level];
for (y = 0; y < h; y++) {
for (x = 0; x < w; x++) {
int v = src[x + y * stride] - ref[x + y * stride];
block[0][x + w * y] = v;
best_score += v * v;
block_sum[0] += v; |
54166165 |
}
}
}
|
ada12f83 |
best_count = 0; |
63a46c61 |
best_score -= (int)((unsigned)block_sum[0] * block_sum[0] >> (level + 3));
best_mean = block_sum[0] + (size >> 1) >> (level + 3); |
54166165 |
|
ada12f83 |
if (level < 4) {
for (count = 1; count < 7; count++) {
int best_vector_score = INT_MAX;
int best_vector_sum = -999, best_vector_mean = -999;
const int stage = count - 1; |
54166165 |
const int8_t *vector;
|
ada12f83 |
for (i = 0; i < 16; i++) {
int sum = codebook_sum[stage * 16 + i]; |
54166165 |
int sqr, diff, score;
|
ada12f83 |
vector = codebook + stage * size * 16 + i * size; |
65d5d586 |
sqr = s->ssd_int8_vs_int16(vector, block[stage], size); |
ada12f83 |
diff = block_sum[stage] - sum; |
41ed7ab4 |
score = sqr - (diff * (int64_t)diff >> (level + 3)); // FIXME: 64 bits slooow |
ada12f83 |
if (score < best_vector_score) { |
63a46c61 |
int mean = diff + (size >> 1) >> (level + 3); |
de31814a |
av_assert2(mean > -300 && mean < 300); |
ada12f83 |
mean = av_clip(mean, intra ? 0 : -256, 255);
best_vector_score = score;
best_vector[stage] = i;
best_vector_sum = sum;
best_vector_mean = mean; |
54166165 |
}
} |
5ee638fd |
av_assert0(best_vector_mean != -999); |
ada12f83 |
vector = codebook + stage * size * 16 + best_vector[stage] * size;
for (j = 0; j < size; j++)
block[stage + 1][j] = block[stage][j] - vector[j];
block_sum[stage + 1] = block_sum[stage] - best_vector_sum;
best_vector_score += lambda *
(+1 + 4 * count +
multistage_vlc[1 + count][1]
+ mean_vlc[best_vector_mean][1]);
if (best_vector_score < best_score) {
best_score = best_vector_score;
best_count = count;
best_mean = best_vector_mean; |
54166165 |
}
}
}
|
ada12f83 |
split = 0;
if (best_score > threshold && level) {
int score = 0; |
63a46c61 |
int offset = level & 1 ? stride * h / 2 : w / 2; |
54166165 |
PutBitContext backup[6];
|
ada12f83 |
for (i = level - 1; i >= 0; i--)
backup[i] = s->reorder_pb[i];
score += encode_block(s, src, ref, decoded, stride, level - 1,
threshold >> 1, lambda, intra);
score += encode_block(s, src + offset, ref + offset, decoded + offset,
stride, level - 1, threshold >> 1, lambda, intra); |
54166165 |
score += lambda;
|
ada12f83 |
if (score < best_score) {
best_score = score;
split = 1;
} else {
for (i = level - 1; i >= 0; i--)
s->reorder_pb[i] = backup[i]; |
54166165 |
}
}
if (level > 0)
put_bits(&s->reorder_pb[level], 1, split);
|
ada12f83 |
if (!split) { |
de31814a |
av_assert1(best_mean >= 0 && best_mean < 256 || !intra);
av_assert1(best_mean >= -256 && best_mean < 256);
av_assert1(best_count >= 0 && best_count < 7);
av_assert1(level < 4 || best_count == 0); |
54166165 |
/* output the encoding */
put_bits(&s->reorder_pb[level], |
ada12f83 |
multistage_vlc[1 + best_count][1],
multistage_vlc[1 + best_count][0]); |
54166165 |
put_bits(&s->reorder_pb[level], mean_vlc[best_mean][1], |
ada12f83 |
mean_vlc[best_mean][0]); |
54166165 |
|
ada12f83 |
for (i = 0; i < best_count; i++) { |
de31814a |
av_assert2(best_vector[i] >= 0 && best_vector[i] < 16); |
54166165 |
put_bits(&s->reorder_pb[level], 4, best_vector[i]);
}
|
ada12f83 |
for (y = 0; y < h; y++)
for (x = 0; x < w; x++)
decoded[x + y * stride] = src[x + y * stride] -
block[best_count][x + w * y] +
best_mean; |
54166165 |
}
return best_score;
}
|
a81ee265 |
static void init_block_index(MpegEncContext *s){
s->block_index[0]= s->b8_stride*(s->mb_y*2 ) + s->mb_x*2;
s->block_index[1]= s->b8_stride*(s->mb_y*2 ) + 1 + s->mb_x*2;
s->block_index[2]= s->b8_stride*(s->mb_y*2 + 1) + s->mb_x*2;
s->block_index[3]= s->b8_stride*(s->mb_y*2 + 1) + 1 + s->mb_x*2;
s->block_index[4]= s->mb_stride*(s->mb_y + 1) + s->b8_stride*s->mb_height*2 + s->mb_x;
s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x;
}
|
2ea2612d |
static int svq1_encode_plane(SVQ1EncContext *s, int plane, |
ada12f83 |
unsigned char *src_plane,
unsigned char *ref_plane,
unsigned char *decoded_plane,
int width, int height, int src_stride, int stride) |
54166165 |
{
int x, y;
int i;
int block_width, block_height;
int level;
int threshold[6]; |
043bcdcd |
uint8_t *src = s->scratchbuf + stride * 32; |
56672aea |
const int lambda = (s->quality * s->quality) >> |
ada12f83 |
(2 * FF_LAMBDA_SHIFT); |
54166165 |
/* figure out the acceptable level thresholds in advance */
threshold[5] = QUALITY_THRESHOLD;
for (level = 4; level >= 0; level--)
threshold[level] = threshold[level + 1] * THRESHOLD_MULTIPLIER;
|
ada12f83 |
block_width = (width + 15) / 16; |
54166165 |
block_height = (height + 15) / 16;
|
56672aea |
if (s->pict_type == AV_PICTURE_TYPE_P) { |
ada12f83 |
s->m.avctx = s->avctx;
s->m.current_picture_ptr = &s->m.current_picture;
s->m.last_picture_ptr = &s->m.last_picture; |
f6774f90 |
s->m.last_picture.f->data[0] = ref_plane; |
ada12f83 |
s->m.linesize = |
f6774f90 |
s->m.last_picture.f->linesize[0] =
s->m.new_picture.f->linesize[0] =
s->m.current_picture.f->linesize[0] = stride; |
ada12f83 |
s->m.width = width;
s->m.height = height;
s->m.mb_width = block_width;
s->m.mb_height = block_height;
s->m.mb_stride = s->m.mb_width + 1;
s->m.b8_stride = 2 * s->m.mb_width + 1;
s->m.f_code = 1; |
56672aea |
s->m.pict_type = s->pict_type; |
4b6b1082 |
s->m.motion_est = s->motion_est; |
ada12f83 |
s->m.me.scene_change_score = 0;
// s->m.out_format = FMT_H263;
// s->m.unrestricted_mv = 1; |
56672aea |
s->m.lambda = s->quality; |
63a46c61 |
s->m.qscale = s->m.lambda * 139 +
FF_LAMBDA_SCALE * 64 >>
FF_LAMBDA_SHIFT + 7;
s->m.lambda2 = s->m.lambda * s->m.lambda +
FF_LAMBDA_SCALE / 2 >> |
ada12f83 |
FF_LAMBDA_SHIFT;
if (!s->motion_val8[plane]) {
s->motion_val8[plane] = av_mallocz((s->m.b8_stride *
block_height * 2 + 2) *
2 * sizeof(int16_t));
s->motion_val16[plane] = av_mallocz((s->m.mb_stride *
(block_height + 2) + 1) *
2 * sizeof(int16_t)); |
2a3af772 |
if (!s->motion_val8[plane] || !s->motion_val16[plane])
return AVERROR(ENOMEM); |
54166165 |
}
|
ada12f83 |
s->m.mb_type = s->mb_type; |
54166165 |
|
ada12f83 |
// dummies, to avoid segfaults
s->m.current_picture.mb_mean = (uint8_t *)s->dummy;
s->m.current_picture.mb_var = (uint16_t *)s->dummy;
s->m.current_picture.mc_mb_var = (uint16_t *)s->dummy; |
759001c5 |
s->m.current_picture.mb_type = s->dummy; |
54166165 |
|
759001c5 |
s->m.current_picture.motion_val[0] = s->motion_val8[plane] + 2; |
ada12f83 |
s->m.p_mv_table = s->motion_val16[plane] +
s->m.mb_stride + 1; |
2d604443 |
s->m.mecc = s->mecc; // move |
54166165 |
ff_init_me(&s->m);
|
ada12f83 |
s->m.me.dia_size = s->avctx->dia_size;
s->m.first_slice_line = 1; |
54166165 |
for (y = 0; y < block_height; y++) { |
f6774f90 |
s->m.new_picture.f->data[0] = src - y * 16 * stride; // ugly |
ada12f83 |
s->m.mb_y = y;
for (i = 0; i < 16 && i + 16 * y < height; i++) {
memcpy(&src[i * stride], &src_plane[(i + 16 * y) * src_stride],
width);
for (x = width; x < 16 * block_width; x++)
src[i * stride + x] = src[i * stride + x - 1]; |
54166165 |
} |
ada12f83 |
for (; i < 16 && i + 16 * y < 16 * block_height; i++)
memcpy(&src[i * stride], &src[(i - 1) * stride],
16 * block_width); |
54166165 |
for (x = 0; x < block_width; x++) { |
ada12f83 |
s->m.mb_x = x; |
a81ee265 |
init_block_index(&s->m); |
54166165 |
ff_estimate_p_frame_motion(&s->m, x, y);
} |
ada12f83 |
s->m.first_slice_line = 0; |
54166165 |
}
ff_fix_long_p_mvs(&s->m); |
ada12f83 |
ff_fix_long_mvs(&s->m, NULL, 0, s->m.p_mv_table, s->m.f_code,
CANDIDATE_MB_TYPE_INTER, 0); |
54166165 |
}
|
ada12f83 |
s->m.first_slice_line = 1; |
54166165 |
for (y = 0; y < block_height; y++) { |
ada12f83 |
for (i = 0; i < 16 && i + 16 * y < height; i++) {
memcpy(&src[i * stride], &src_plane[(i + 16 * y) * src_stride],
width);
for (x = width; x < 16 * block_width; x++)
src[i * stride + x] = src[i * stride + x - 1]; |
54166165 |
} |
ada12f83 |
for (; i < 16 && i + 16 * y < 16 * block_height; i++)
memcpy(&src[i * stride], &src[(i - 1) * stride], 16 * block_width); |
54166165 |
|
ada12f83 |
s->m.mb_y = y; |
54166165 |
for (x = 0; x < block_width; x++) { |
fcfc90ed |
uint8_t reorder_buffer[2][6][7 * 32];
int count[2][6]; |
ada12f83 |
int offset = y * 16 * stride + x * 16;
uint8_t *decoded = decoded_plane + offset;
uint8_t *ref = ref_plane + offset;
int score[4] = { 0, 0, 0, 0 }, best;
uint8_t *temp = s->scratchbuf;
if (s->pb.buf_end - s->pb.buf -
(put_bits_count(&s->pb) >> 3) < 3000) { // FIXME: check size |
54166165 |
av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
return -1;
}
|
ada12f83 |
s->m.mb_x = x; |
a81ee265 |
init_block_index(&s->m); |
54166165 |
|
56672aea |
if (s->pict_type == AV_PICTURE_TYPE_I || |
ada12f83 |
(s->m.mb_type[x + y * s->m.mb_stride] &
CANDIDATE_MB_TYPE_INTRA)) {
for (i = 0; i < 6; i++)
init_put_bits(&s->reorder_pb[i], reorder_buffer[0][i],
7 * 32); |
56672aea |
if (s->pict_type == AV_PICTURE_TYPE_P) { |
ada12f83 |
const uint8_t *vlc = ff_svq1_block_type_vlc[SVQ1_BLOCK_INTRA]; |
54166165 |
put_bits(&s->reorder_pb[5], vlc[1], vlc[0]); |
ada12f83 |
score[0] = vlc[1] * lambda; |
54166165 |
} |
ada12f83 |
score[0] += encode_block(s, src + 16 * x, NULL, temp, stride,
5, 64, lambda, 1);
for (i = 0; i < 6; i++) {
count[0][i] = put_bits_count(&s->reorder_pb[i]); |
54166165 |
flush_put_bits(&s->reorder_pb[i]);
} |
ada12f83 |
} else
score[0] = INT_MAX; |
54166165 |
|
ada12f83 |
best = 0; |
54166165 |
|
56672aea |
if (s->pict_type == AV_PICTURE_TYPE_P) { |
ada12f83 |
const uint8_t *vlc = ff_svq1_block_type_vlc[SVQ1_BLOCK_INTER]; |
54166165 |
int mx, my, pred_x, pred_y, dxy;
int16_t *motion_ptr;
|
ada12f83 |
motion_ptr = ff_h263_pred_motion(&s->m, 0, 0, &pred_x, &pred_y);
if (s->m.mb_type[x + y * s->m.mb_stride] &
CANDIDATE_MB_TYPE_INTER) {
for (i = 0; i < 6; i++)
init_put_bits(&s->reorder_pb[i], reorder_buffer[1][i],
7 * 32); |
54166165 |
put_bits(&s->reorder_pb[5], vlc[1], vlc[0]);
|
ada12f83 |
s->m.pb = s->reorder_pb[5];
mx = motion_ptr[0];
my = motion_ptr[1]; |
de31814a |
av_assert1(mx >= -32 && mx <= 31);
av_assert1(my >= -32 && my <= 31);
av_assert1(pred_x >= -32 && pred_x <= 31);
av_assert1(pred_y >= -32 && pred_y <= 31); |
404fe63e |
ff_h263_encode_motion(&s->m.pb, mx - pred_x, 1);
ff_h263_encode_motion(&s->m.pb, my - pred_y, 1); |
ada12f83 |
s->reorder_pb[5] = s->m.pb;
score[1] += lambda * put_bits_count(&s->reorder_pb[5]);
dxy = (mx & 1) + 2 * (my & 1);
|
043bcdcd |
s->hdsp.put_pixels_tab[0][dxy](temp + 16*stride, |
1277dc07 |
ref + (mx >> 1) +
stride * (my >> 1),
stride, 16); |
ada12f83 |
|
043bcdcd |
score[1] += encode_block(s, src + 16 * x, temp + 16*stride, |
ada12f83 |
decoded, stride, 5, 64, lambda, 0);
best = score[1] <= score[0];
vlc = ff_svq1_block_type_vlc[SVQ1_BLOCK_SKIP]; |
2d604443 |
score[2] = s->mecc.sse[0](NULL, src + 16 * x, ref,
stride, 16); |
ada12f83 |
score[2] += vlc[1] * lambda;
if (score[2] < score[best] && mx == 0 && my == 0) {
best = 2; |
b42d594c |
s->hdsp.put_pixels_tab[0][0](decoded, ref, stride, 16); |
54166165 |
put_bits(&s->pb, vlc[1], vlc[0]);
}
}
|
ada12f83 |
if (best == 1) {
for (i = 0; i < 6; i++) {
count[1][i] = put_bits_count(&s->reorder_pb[i]); |
54166165 |
flush_put_bits(&s->reorder_pb[i]);
} |
ada12f83 |
} else {
motion_ptr[0] =
motion_ptr[1] =
motion_ptr[2] =
motion_ptr[3] =
motion_ptr[0 + 2 * s->m.b8_stride] =
motion_ptr[1 + 2 * s->m.b8_stride] =
motion_ptr[2 + 2 * s->m.b8_stride] =
motion_ptr[3 + 2 * s->m.b8_stride] = 0; |
54166165 |
}
}
s->rd_total += score[best];
|
fcfc90ed |
if (best != 2) |
ada12f83 |
for (i = 5; i >= 0; i--)
avpriv_copy_bits(&s->pb, reorder_buffer[best][i],
count[best][i]);
if (best == 0) |
b42d594c |
s->hdsp.put_pixels_tab[0][0](decoded, temp, stride, 16); |
54166165 |
} |
ada12f83 |
s->m.first_slice_line = 0; |
54166165 |
}
return 0;
}
|
394ef4d1 |
static av_cold int svq1_encode_end(AVCodecContext *avctx)
{ |
2ea2612d |
SVQ1EncContext *const s = avctx->priv_data; |
394ef4d1 |
int i;
av_log(avctx, AV_LOG_DEBUG, "RD: %f\n",
s->rd_total / (double)(avctx->width * avctx->height *
avctx->frame_number));
|
f6774f90 |
s->m.mb_type = NULL; |
835f798c |
ff_mpv_common_end(&s->m); |
f6774f90 |
|
394ef4d1 |
av_freep(&s->m.me.scratchpad);
av_freep(&s->m.me.map);
av_freep(&s->m.me.score_map);
av_freep(&s->mb_type);
av_freep(&s->dummy);
av_freep(&s->scratchbuf);
for (i = 0; i < 3; i++) {
av_freep(&s->motion_val8[i]);
av_freep(&s->motion_val16[i]);
}
av_frame_free(&s->current_picture);
av_frame_free(&s->last_picture);
return 0;
}
|
98a6fff9 |
static av_cold int svq1_encode_init(AVCodecContext *avctx) |
54166165 |
{ |
2ea2612d |
SVQ1EncContext *const s = avctx->priv_data; |
f6774f90 |
int ret; |
54166165 |
|
88fe45e0 |
if (avctx->width >= 4096 || avctx->height >= 4096) {
av_log(avctx, AV_LOG_ERROR, "Dimensions too large, maximum is 4095x4095\n");
return AVERROR(EINVAL);
}
|
b42d594c |
ff_hpeldsp_init(&s->hdsp, avctx->flags); |
2d604443 |
ff_me_cmp_init(&s->mecc, avctx); |
c1661484 |
ff_mpegvideoencdsp_init(&s->m.mpvencdsp, avctx); |
394ef4d1 |
s->current_picture = av_frame_alloc();
s->last_picture = av_frame_alloc(); |
d6604b29 |
if (!s->current_picture || !s->last_picture) { |
394ef4d1 |
svq1_encode_end(avctx);
return AVERROR(ENOMEM);
} |
54166165 |
|
ada12f83 |
s->frame_width = avctx->width; |
54166165 |
s->frame_height = avctx->height;
|
ada12f83 |
s->y_block_width = (s->frame_width + 15) / 16; |
54166165 |
s->y_block_height = (s->frame_height + 15) / 16;
|
ada12f83 |
s->c_block_width = (s->frame_width / 4 + 15) / 16; |
54166165 |
s->c_block_height = (s->frame_height / 4 + 15) / 16;
|
ada12f83 |
s->avctx = avctx;
s->m.avctx = avctx; |
f6774f90 |
|
835f798c |
if ((ret = ff_mpv_common_init(&s->m)) < 0) { |
f6774f90 |
svq1_encode_end(avctx);
return ret;
}
|
3e0b4e32 |
s->m.picture_structure = PICT_FRAME; |
ada12f83 |
s->m.me.temp =
s->m.me.scratchpad = av_mallocz((avctx->width + 64) *
2 * 16 * 2 * sizeof(uint8_t));
s->m.me.map = av_mallocz(ME_MAP_SIZE * sizeof(uint32_t));
s->m.me.score_map = av_mallocz(ME_MAP_SIZE * sizeof(uint32_t));
s->mb_type = av_mallocz((s->y_block_width + 1) *
s->y_block_height * sizeof(int16_t));
s->dummy = av_mallocz((s->y_block_width + 1) *
s->y_block_height * sizeof(int32_t)); |
65d5d586 |
s->ssd_int8_vs_int16 = ssd_int8_vs_int16_c;
|
28fb80dc |
if (!s->m.me.temp || !s->m.me.scratchpad || !s->m.me.map ||
!s->m.me.score_map || !s->mb_type || !s->dummy) { |
2a3af772 |
svq1_encode_end(avctx);
return AVERROR(ENOMEM);
}
|
65d5d586 |
if (ARCH_PPC)
ff_svq1enc_init_ppc(s);
if (ARCH_X86)
ff_svq1enc_init_x86(s);
|
ada12f83 |
ff_h263_encode_init(&s->m); // mv_penalty |
54166165 |
return 0;
}
|
3cc77ded |
static int svq1_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
const AVFrame *pict, int *got_packet) |
54166165 |
{ |
2ea2612d |
SVQ1EncContext *const s = avctx->priv_data; |
3cc77ded |
int i, ret;
|
de31814a |
if ((ret = ff_alloc_packet2(avctx, pkt, s->y_block_width * s->y_block_height * |
29d147c9 |
MAX_MB_BYTES*3 + AV_INPUT_BUFFER_MIN_SIZE, 0)) < 0) |
3cc77ded |
return ret; |
54166165 |
|
ada12f83 |
if (avctx->pix_fmt != AV_PIX_FMT_YUV410P) { |
54166165 |
av_log(avctx, AV_LOG_ERROR, "unsupported pixel format\n");
return -1;
}
|
394ef4d1 |
if (!s->current_picture->data[0]) { |
79888388 |
if ((ret = ff_get_buffer(avctx, s->current_picture, 0)) < 0) { |
7effc26b |
return ret;
} |
54166165 |
} |
79888388 |
if (!s->last_picture->data[0]) { |
59846452 |
ret = ff_get_buffer(avctx, s->last_picture, 0);
if (ret < 0) |
79888388 |
return ret;
}
if (!s->scratchbuf) {
s->scratchbuf = av_malloc_array(s->current_picture->linesize[0], 16 * 3);
if (!s->scratchbuf)
return AVERROR(ENOMEM);
} |
54166165 |
|
394ef4d1 |
FFSWAP(AVFrame*, s->current_picture, s->last_picture); |
54166165 |
|
3cc77ded |
init_put_bits(&s->pb, pkt->data, pkt->size); |
54166165 |
|
56672aea |
if (avctx->gop_size && (avctx->frame_number % avctx->gop_size))
s->pict_type = AV_PICTURE_TYPE_P;
else
s->pict_type = AV_PICTURE_TYPE_I;
s->quality = pict->quality; |
54166165 |
|
40cf1bba |
#if FF_API_CODED_FRAME
FF_DISABLE_DEPRECATION_WARNINGS |
56672aea |
avctx->coded_frame->pict_type = s->pict_type;
avctx->coded_frame->key_frame = s->pict_type == AV_PICTURE_TYPE_I; |
40cf1bba |
FF_ENABLE_DEPRECATION_WARNINGS
#endif |
56672aea |
|
c40ecffd |
ff_side_data_set_encoder_stats(pkt, pict->quality, NULL, 0, s->pict_type); |
5d3addb9 |
|
56672aea |
svq1_write_header(s, s->pict_type); |
70dc6bbf |
for (i = 0; i < 3; i++) {
int ret = svq1_encode_plane(s, i, |
394ef4d1 |
pict->data[i],
s->last_picture->data[i],
s->current_picture->data[i], |
ada12f83 |
s->frame_width / (i ? 4 : 1),
s->frame_height / (i ? 4 : 1), |
394ef4d1 |
pict->linesize[i], |
70dc6bbf |
s->current_picture->linesize[i]);
emms_c();
if (ret < 0) { |
28fb80dc |
int j;
for (j = 0; j < i; j++) {
av_freep(&s->motion_val8[j]);
av_freep(&s->motion_val16[j]);
}
av_freep(&s->scratchbuf); |
ada12f83 |
return -1; |
28fb80dc |
} |
70dc6bbf |
} |
ada12f83 |
// avpriv_align_put_bits(&s->pb);
while (put_bits_count(&s->pb) & 31) |
54166165 |
put_bits(&s->pb, 1, 0);
flush_put_bits(&s->pb);
|
3cc77ded |
pkt->size = put_bits_count(&s->pb) / 8; |
56672aea |
if (s->pict_type == AV_PICTURE_TYPE_I) |
3cc77ded |
pkt->flags |= AV_PKT_FLAG_KEY;
*got_packet = 1;
return 0; |
54166165 |
}
|
4b6b1082 |
#define OFFSET(x) offsetof(struct SVQ1EncContext, x)
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
{ "motion-est", "Motion estimation algorithm", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_XONE, VE, "motion-est"},
{ "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, "motion-est" },
{ "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, "motion-est" },
{ "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, "motion-est" },
{ NULL },
};
static const AVClass svq1enc_class = {
.class_name = "svq1enc",
.item_name = av_default_item_name,
.option = options,
.version = LIBAVUTIL_VERSION_INT,
};
|
e7e2df27 |
AVCodec ff_svq1_encoder = { |
ec6402b7 |
.name = "svq1", |
b2bed932 |
.long_name = NULL_IF_CONFIG_SMALL("Sorenson Vector Quantizer 1 / Sorenson Video 1 / SVQ1"), |
ec6402b7 |
.type = AVMEDIA_TYPE_VIDEO, |
36ef5369 |
.id = AV_CODEC_ID_SVQ1, |
2ea2612d |
.priv_data_size = sizeof(SVQ1EncContext), |
4b6b1082 |
.priv_class = &svq1enc_class, |
ec6402b7 |
.init = svq1_encode_init, |
3cc77ded |
.encode2 = svq1_encode_frame, |
ec6402b7 |
.close = svq1_encode_end, |
542b83fc |
.pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV410P,
AV_PIX_FMT_NONE }, |
54166165 |
}; |