blob: b6ae92b2ac79d43ac68c378c4c264a95800aaf4f
1 | /* |
2 | * Microsoft Video-1 Encoder |
3 | * Copyright (c) 2009 Konstantin Shishkov |
4 | * |
5 | * This file is part of FFmpeg. |
6 | * |
7 | * FFmpeg is free software; you can redistribute it and/or |
8 | * modify it under the terms of the GNU Lesser General Public |
9 | * License as published by the Free Software Foundation; either |
10 | * version 2.1 of the License, or (at your option) any later version. |
11 | * |
12 | * FFmpeg is distributed in the hope that it will be useful, |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
15 | * Lesser General Public License for more details. |
16 | * |
17 | * You should have received a copy of the GNU Lesser General Public |
18 | * License along with FFmpeg; if not, write to the Free Software |
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 | */ |
21 | |
22 | /** |
23 | * @file |
24 | * Microsoft Video-1 encoder |
25 | */ |
26 | |
27 | #include "avcodec.h" |
28 | #include "internal.h" |
29 | #include "bytestream.h" |
30 | #include "libavutil/lfg.h" |
31 | #include "elbg.h" |
32 | #include "libavutil/imgutils.h" |
33 | /** |
34 | * Encoder context |
35 | */ |
36 | typedef struct Msvideo1EncContext { |
37 | AVCodecContext *avctx; |
38 | AVLFG rnd; |
39 | uint8_t *prev; |
40 | |
41 | int block[16*3]; |
42 | int block2[16*3]; |
43 | int codebook[8*3]; |
44 | int codebook2[8*3]; |
45 | int output[16*3]; |
46 | int output2[16*3]; |
47 | int avg[3]; |
48 | int bestpos; |
49 | int keyint; |
50 | } Msvideo1EncContext; |
51 | |
52 | enum MSV1Mode{ |
53 | MODE_SKIP = 0, |
54 | MODE_FILL, |
55 | MODE_2COL, |
56 | MODE_8COL, |
57 | }; |
58 | |
59 | #define SKIP_PREFIX 0x8400 |
60 | #define SKIPS_MAX 0x03FF |
61 | #define MKRGB555(in, off) (((in)[off] << 10) | ((in)[(off) + 1] << 5) | ((in)[(off) + 2])) |
62 | |
63 | static const int remap[16] = { 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15 }; |
64 | |
65 | static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, |
66 | const AVFrame *pict, int *got_packet) |
67 | { |
68 | Msvideo1EncContext * const c = avctx->priv_data; |
69 | const AVFrame *p = pict; |
70 | uint16_t *src; |
71 | uint8_t *prevptr; |
72 | uint8_t *dst, *buf; |
73 | int keyframe = 0; |
74 | int no_skips = 1; |
75 | int i, j, k, x, y, ret; |
76 | int skips = 0; |
77 | int quality = 24; |
78 | |
79 | if ((ret = ff_alloc_packet2(avctx, pkt, avctx->width*avctx->height*9 + AV_INPUT_BUFFER_MIN_SIZE, 0)) < 0) |
80 | return ret; |
81 | dst= buf= pkt->data; |
82 | |
83 | if(!c->prev) |
84 | c->prev = av_malloc(avctx->width * 3 * (avctx->height + 3)); |
85 | prevptr = c->prev + avctx->width * 3 * (FFALIGN(avctx->height, 4) - 1); |
86 | src = (uint16_t*)(p->data[0] + p->linesize[0]*(FFALIGN(avctx->height, 4) - 1)); |
87 | if(c->keyint >= avctx->keyint_min) |
88 | keyframe = 1; |
89 | |
90 | |
91 | for(y = 0; y < avctx->height; y += 4){ |
92 | for(x = 0; x < avctx->width; x += 4){ |
93 | int bestmode = MODE_SKIP; |
94 | int bestscore = INT_MAX; |
95 | int flags = 0; |
96 | int score; |
97 | |
98 | for(j = 0; j < 4; j++){ |
99 | for(i = 0; i < 4; i++){ |
100 | uint16_t val = src[x + i - j*p->linesize[0]/2]; |
101 | for(k = 0; k < 3; k++){ |
102 | c->block[(i + j*4)*3 + k] = |
103 | c->block2[remap[i + j*4]*3 + k] = (val >> (10-k*5)) & 0x1F; |
104 | } |
105 | } |
106 | } |
107 | if(!keyframe){ |
108 | bestscore = 0; |
109 | for(j = 0; j < 4; j++){ |
110 | for(i = 0; i < 4*3; i++){ |
111 | int t = prevptr[x*3 + i - j*3*avctx->width] - c->block[i + j*4*3]; |
112 | bestscore += t*t; |
113 | } |
114 | } |
115 | bestscore /= quality; |
116 | } |
117 | // try to find optimal value to fill whole 4x4 block |
118 | score = 0; |
119 | avpriv_init_elbg(c->block, 3, 16, c->avg, 1, 1, c->output, &c->rnd); |
120 | avpriv_do_elbg (c->block, 3, 16, c->avg, 1, 1, c->output, &c->rnd); |
121 | if(c->avg[0] == 1) // red component = 1 will be written as skip code |
122 | c->avg[0] = 0; |
123 | for(j = 0; j < 4; j++){ |
124 | for(i = 0; i < 4; i++){ |
125 | for(k = 0; k < 3; k++){ |
126 | int t = c->avg[k] - c->block[(i+j*4)*3+k]; |
127 | score += t*t; |
128 | } |
129 | } |
130 | } |
131 | score /= quality; |
132 | score += 2; |
133 | if(score < bestscore){ |
134 | bestscore = score; |
135 | bestmode = MODE_FILL; |
136 | } |
137 | // search for optimal filling of 2-color block |
138 | score = 0; |
139 | avpriv_init_elbg(c->block, 3, 16, c->codebook, 2, 1, c->output, &c->rnd); |
140 | avpriv_do_elbg (c->block, 3, 16, c->codebook, 2, 1, c->output, &c->rnd); |
141 | // last output value should be always 1, swap codebooks if needed |
142 | if(!c->output[15]){ |
143 | for(i = 0; i < 3; i++) |
144 | FFSWAP(uint8_t, c->codebook[i], c->codebook[i+3]); |
145 | for(i = 0; i < 16; i++) |
146 | c->output[i] ^= 1; |
147 | } |
148 | for(j = 0; j < 4; j++){ |
149 | for(i = 0; i < 4; i++){ |
150 | for(k = 0; k < 3; k++){ |
151 | int t = c->codebook[c->output[i+j*4]*3 + k] - c->block[i*3+k+j*4*3]; |
152 | score += t*t; |
153 | } |
154 | } |
155 | } |
156 | score /= quality; |
157 | score += 6; |
158 | if(score < bestscore){ |
159 | bestscore = score; |
160 | bestmode = MODE_2COL; |
161 | } |
162 | // search for optimal filling of 2-color 2x2 subblocks |
163 | score = 0; |
164 | for(i = 0; i < 4; i++){ |
165 | avpriv_init_elbg(c->block2 + i*4*3, 3, 4, c->codebook2 + i*2*3, 2, 1, c->output2 + i*4, &c->rnd); |
166 | avpriv_do_elbg (c->block2 + i*4*3, 3, 4, c->codebook2 + i*2*3, 2, 1, c->output2 + i*4, &c->rnd); |
167 | } |
168 | // last value should be always 1, swap codebooks if needed |
169 | if(!c->output2[15]){ |
170 | for(i = 0; i < 3; i++) |
171 | FFSWAP(uint8_t, c->codebook2[i+18], c->codebook2[i+21]); |
172 | for(i = 12; i < 16; i++) |
173 | c->output2[i] ^= 1; |
174 | } |
175 | for(j = 0; j < 4; j++){ |
176 | for(i = 0; i < 4; i++){ |
177 | for(k = 0; k < 3; k++){ |
178 | int t = c->codebook2[(c->output2[remap[i+j*4]] + (i&2) + (j&2)*2)*3+k] - c->block[i*3+k + j*4*3]; |
179 | score += t*t; |
180 | } |
181 | } |
182 | } |
183 | score /= quality; |
184 | score += 18; |
185 | if(score < bestscore){ |
186 | bestscore = score; |
187 | bestmode = MODE_8COL; |
188 | } |
189 | |
190 | if(bestmode == MODE_SKIP){ |
191 | skips++; |
192 | no_skips = 0; |
193 | } |
194 | if((bestmode != MODE_SKIP && skips) || skips == SKIPS_MAX){ |
195 | bytestream_put_le16(&dst, skips | SKIP_PREFIX); |
196 | skips = 0; |
197 | } |
198 | |
199 | switch(bestmode){ |
200 | case MODE_FILL: |
201 | bytestream_put_le16(&dst, MKRGB555(c->avg,0) | 0x8000); |
202 | for(j = 0; j < 4; j++) |
203 | for(i = 0; i < 4; i++) |
204 | for(k = 0; k < 3; k++) |
205 | prevptr[x*3 + i*3 + k - j*3*avctx->width] = c->avg[k]; |
206 | break; |
207 | case MODE_2COL: |
208 | for(j = 0; j < 4; j++){ |
209 | for(i = 0; i < 4; i++){ |
210 | flags |= (c->output[i + j*4]^1) << (i + j*4); |
211 | for(k = 0; k < 3; k++) |
212 | prevptr[x*3 + i*3 + k - j*3*avctx->width] = c->codebook[c->output[i + j*4]*3 + k]; |
213 | } |
214 | } |
215 | bytestream_put_le16(&dst, flags); |
216 | bytestream_put_le16(&dst, MKRGB555(c->codebook, 0)); |
217 | bytestream_put_le16(&dst, MKRGB555(c->codebook, 3)); |
218 | break; |
219 | case MODE_8COL: |
220 | for(j = 0; j < 4; j++){ |
221 | for(i = 0; i < 4; i++){ |
222 | flags |= (c->output2[remap[i + j*4]]^1) << (i + j*4); |
223 | for(k = 0; k < 3; k++) |
224 | prevptr[x*3 + i*3 + k - j*3*avctx->width] = c->codebook2[(c->output2[remap[i+j*4]] + (i&2) + (j&2)*2)*3 + k]; |
225 | } |
226 | } |
227 | bytestream_put_le16(&dst, flags); |
228 | bytestream_put_le16(&dst, MKRGB555(c->codebook2, 0) | 0x8000); |
229 | for(i = 3; i < 24; i += 3) |
230 | bytestream_put_le16(&dst, MKRGB555(c->codebook2, i)); |
231 | break; |
232 | } |
233 | } |
234 | src -= p->linesize[0] << 1; |
235 | prevptr -= avctx->width * 3 * 4; |
236 | } |
237 | if(skips) |
238 | bytestream_put_le16(&dst, skips | SKIP_PREFIX); |
239 | //EOF |
240 | bytestream_put_byte(&dst, 0); |
241 | bytestream_put_byte(&dst, 0); |
242 | |
243 | if(no_skips) |
244 | keyframe = 1; |
245 | if(keyframe) |
246 | c->keyint = 0; |
247 | else |
248 | c->keyint++; |
249 | if (keyframe) pkt->flags |= AV_PKT_FLAG_KEY; |
250 | pkt->size = dst - buf; |
251 | *got_packet = 1; |
252 | |
253 | return 0; |
254 | } |
255 | |
256 | |
257 | /** |
258 | * init encoder |
259 | */ |
260 | static av_cold int encode_init(AVCodecContext *avctx) |
261 | { |
262 | Msvideo1EncContext * const c = avctx->priv_data; |
263 | |
264 | c->avctx = avctx; |
265 | if (av_image_check_size(avctx->width, avctx->height, 0, avctx) < 0) { |
266 | return -1; |
267 | } |
268 | if((avctx->width&3) || (avctx->height&3)){ |
269 | av_log(avctx, AV_LOG_ERROR, "width and height must be multiples of 4\n"); |
270 | return -1; |
271 | } |
272 | |
273 | avctx->bits_per_coded_sample = 16; |
274 | |
275 | c->keyint = avctx->keyint_min; |
276 | av_lfg_init(&c->rnd, 1); |
277 | |
278 | return 0; |
279 | } |
280 | |
281 | |
282 | |
283 | /** |
284 | * Uninit encoder |
285 | */ |
286 | static av_cold int encode_end(AVCodecContext *avctx) |
287 | { |
288 | Msvideo1EncContext * const c = avctx->priv_data; |
289 | |
290 | av_freep(&c->prev); |
291 | |
292 | return 0; |
293 | } |
294 | |
295 | AVCodec ff_msvideo1_encoder = { |
296 | .name = "msvideo1", |
297 | .long_name = NULL_IF_CONFIG_SMALL("Microsoft Video-1"), |
298 | .type = AVMEDIA_TYPE_VIDEO, |
299 | .id = AV_CODEC_ID_MSVIDEO1, |
300 | .priv_data_size = sizeof(Msvideo1EncContext), |
301 | .init = encode_init, |
302 | .encode2 = encode_frame, |
303 | .close = encode_end, |
304 | .pix_fmts = (const enum AVPixelFormat[]){AV_PIX_FMT_RGB555, AV_PIX_FMT_NONE}, |
305 | }; |
306 |