Commit | Line | Data |
---|---|---|
a178c22c KS |
1 | use nihav_core::codecs::*; |
2 | use nihav_core::io::byteio::*; | |
3 | use nihav_codec_support::vq::*; | |
4 | ||
5 | #[derive(Default,Clone,Copy,PartialEq,Debug)] | |
6 | struct YUVCode { | |
7 | y: [u8; 4], | |
8 | u: u8, | |
9 | v: u8, | |
10 | } | |
11 | impl VQElement for YUVCode { | |
12 | fn dist(&self, rval: Self) -> u32 { | |
13 | let mut ysum = 0; | |
14 | for (y0, y1) in self.y.iter().zip(rval.y.iter()) { | |
15 | let yd = i32::from(*y0) - i32::from(*y1); | |
16 | ysum += yd * yd; | |
17 | } | |
18 | let ud = i32::from(self.u) - i32::from(rval.u); | |
19 | let vd = i32::from(self.v) - i32::from(rval.v); | |
20 | (ysum + ud * ud + vd * vd) as u32 | |
21 | } | |
22 | fn min_cw() -> Self { YUVCode { y: [0; 4], u: 0, v: 0 } } | |
23 | fn max_cw() -> Self { YUVCode { y: [255; 4], u: 255, v: 255 } } | |
24 | fn min(&self, rval: Self) -> Self { | |
25 | let mut ycode = YUVCode::default(); | |
26 | for i in 0..4 { | |
27 | ycode.y[i] = self.y[i].min(rval.y[i]); | |
28 | } | |
29 | ycode.u = self.u.min(rval.u); | |
30 | ycode.v = self.v.min(rval.v); | |
31 | ycode | |
32 | } | |
33 | fn max(&self, rval: Self) -> Self { | |
34 | let mut ycode = YUVCode::default(); | |
35 | for i in 0..4 { | |
36 | ycode.y[i] = self.y[i].max(rval.y[i]); | |
37 | } | |
38 | ycode.u = self.u.max(rval.u); | |
39 | ycode.v = self.v.max(rval.v); | |
40 | ycode | |
41 | } | |
42 | fn num_components() -> usize { 6 } | |
43 | fn sort_by_component(arr: &mut [Self], component: usize) { | |
44 | let mut counts = [0; 256]; | |
45 | for entry in arr.iter() { | |
46 | let idx = match component { | |
47 | 0 | 1 | 2 | 3 => entry.y[component], | |
48 | 4 => entry.u, | |
49 | _ => entry.v, | |
50 | } as usize; | |
51 | counts[idx] += 1; | |
52 | } | |
53 | let mut offs = [0; 256]; | |
54 | for i in 0..255 { | |
55 | offs[i + 1] = offs[i] + counts[i]; | |
56 | } | |
57 | let mut dst = vec![YUVCode::default(); arr.len()]; | |
58 | for entry in arr.iter() { | |
59 | let idx = match component { | |
60 | 0 | 1 | 2 | 3 => entry.y[component], | |
61 | 4 => entry.u, | |
62 | _ => entry.v, | |
63 | } as usize; | |
64 | dst[offs[idx]] = *entry; | |
65 | offs[idx] += 1; | |
66 | } | |
67 | arr.copy_from_slice(dst.as_slice()); | |
68 | } | |
69 | fn max_dist_component(min: &Self, max: &Self) -> usize { | |
70 | let mut comp = 0; | |
71 | let mut diff = 0; | |
72 | for i in 0..4 { | |
73 | let d = u32::from(max.y[i]) - u32::from(min.y[i]); | |
74 | if d > diff { | |
75 | diff = d; | |
76 | comp = i; | |
77 | } | |
78 | } | |
79 | let ud = u32::from(max.u) - u32::from(min.u); | |
80 | if ud > diff { | |
81 | diff = ud; | |
82 | comp = 4; | |
83 | } | |
84 | let vd = u32::from(max.v) - u32::from(min.v); | |
85 | if vd > diff { | |
86 | comp = 5; | |
87 | } | |
88 | comp | |
89 | } | |
90 | } | |
91 | ||
92 | #[derive(Default)] | |
93 | struct YUVCodeSum { | |
94 | ysum: [u64; 4], | |
95 | usum: u64, | |
96 | vsum: u64, | |
97 | count: u64, | |
98 | } | |
99 | ||
100 | impl VQElementSum<YUVCode> for YUVCodeSum { | |
101 | fn zero() -> Self { Self::default() } | |
102 | fn add(&mut self, rval: YUVCode, count: u64) { | |
103 | for i in 0..4 { | |
104 | self.ysum[i] += u64::from(rval.y[i]) * count; | |
105 | } | |
106 | self.usum += u64::from(rval.u) * count; | |
107 | self.vsum += u64::from(rval.v) * count; | |
108 | self.count += count; | |
109 | } | |
110 | fn get_centroid(&self) -> YUVCode { | |
111 | if self.count != 0 { | |
112 | let mut ycode = YUVCode::default(); | |
113 | for i in 0..4 { | |
114 | ycode.y[i] = ((self.ysum[i] + self.count / 2) / self.count) as u8; | |
115 | } | |
116 | ycode.u = ((self.usum + self.count / 2) / self.count) as u8; | |
117 | ycode.v = ((self.vsum + self.count / 2) / self.count) as u8; | |
118 | ycode | |
119 | } else { | |
120 | YUVCode::default() | |
121 | } | |
122 | } | |
123 | } | |
124 | ||
125 | struct RNG { | |
126 | seed: u32, | |
127 | } | |
128 | ||
129 | impl RNG { | |
130 | fn new() -> Self { Self { seed: 0x12345678 } } | |
131 | fn next(&mut self) -> u8 { | |
132 | let mut x = self.seed; | |
133 | x ^= x.wrapping_shl(13); | |
134 | x ^= x >> 17; | |
135 | self.seed = x; | |
136 | (self.seed >> 24) as u8 | |
137 | } | |
138 | fn fill_entry(&mut self, entry: &mut YUVCode) { | |
139 | for y in entry.y.iter_mut() { | |
140 | *y = self.next(); | |
141 | } | |
142 | entry.u = self.next(); | |
143 | entry.v = self.next(); | |
144 | } | |
145 | } | |
146 | ||
147 | const GRAY_FORMAT: NAPixelFormaton = NAPixelFormaton { | |
148 | model: ColorModel::YUV(YUVSubmodel::YUVJ), | |
149 | components: 1, | |
150 | comp_info: [Some(NAPixelChromaton{h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 0, next_elem: 1}), None, None, None, None], | |
151 | elem_size: 1, | |
152 | be: true, | |
153 | alpha: false, | |
154 | palette: false, | |
155 | }; | |
156 | ||
157 | struct MaskWriter { | |
158 | masks: Vec<u32>, | |
159 | mask: u32, | |
160 | pos: u8, | |
161 | } | |
162 | ||
163 | impl MaskWriter { | |
164 | fn new() -> Self { | |
165 | Self { | |
166 | masks: Vec::new(), | |
167 | mask: 0, | |
168 | pos: 0, | |
169 | } | |
170 | } | |
171 | fn reset(&mut self) { | |
172 | self.masks.truncate(0); | |
173 | self.mask = 0; | |
174 | self.pos = 0; | |
175 | } | |
176 | fn put_v1(&mut self) { | |
177 | self.mask <<= 1; | |
178 | self.pos += 1; | |
179 | if self.pos == 32 { | |
180 | self.flush(); | |
181 | } | |
182 | } | |
183 | fn put_v4(&mut self) { | |
184 | self.mask <<= 1; | |
185 | self.mask |= 1; | |
186 | self.pos += 1; | |
187 | if self.pos == 32 { | |
188 | self.flush(); | |
189 | } | |
190 | } | |
191 | fn put_inter(&mut self, skip: bool) { | |
192 | self.mask <<= 1; | |
193 | self.mask |= !skip as u32; | |
194 | self.pos += 1; | |
195 | if self.pos == 32 { | |
196 | self.flush(); | |
197 | } | |
198 | } | |
199 | fn flush(&mut self) { | |
200 | self.masks.push(self.mask); | |
201 | self.mask = 0; | |
202 | self.pos = 0; | |
203 | } | |
204 | fn end(&mut self) { | |
205 | if self.pos == 0 { return; } | |
206 | while self.pos < 32 { | |
207 | self.mask <<= 1; | |
208 | self.pos += 1; | |
209 | } | |
210 | self.flush(); | |
211 | } | |
212 | } | |
213 | ||
214 | struct CinepakEncoder { | |
215 | stream: Option<NAStreamRef>, | |
216 | lastfrm: Option<NAVideoBufferRef<u8>>, | |
217 | pkt: Option<NAPacket>, | |
218 | frmcount: u8, | |
219 | quality: u8, | |
220 | nstrips: usize, | |
221 | v1_entries: Vec<YUVCode>, | |
222 | v4_entries: Vec<YUVCode>, | |
223 | v1_cb: [YUVCode; 256], | |
224 | v4_cb: [YUVCode; 256], | |
225 | v1_cur_cb: [YUVCode; 256], | |
226 | v4_cur_cb: [YUVCode; 256], | |
227 | v1_idx: Vec<u8>, | |
228 | v4_idx: Vec<u8>, | |
229 | grayscale: bool, | |
230 | rng: RNG, | |
231 | masks: MaskWriter, | |
232 | skip_dist: Vec<u32>, | |
233 | } | |
234 | ||
235 | fn avg4(a: u8, b: u8, c: u8, d: u8) -> u8 { | |
236 | ((u16::from(a) + u16::from(b) + u16::from(c) + u16::from(d) + 3) >> 2) as u8 | |
237 | } | |
238 | ||
239 | fn patch_size(bw: &mut ByteWriter, pos: u64) -> EncoderResult<()> { | |
240 | let size = bw.tell() - pos; | |
241 | bw.seek(SeekFrom::Current(-((size + 3) as i64)))?; | |
242 | bw.write_u24be((size + 4) as u32)?; | |
243 | bw.seek(SeekFrom::End(0))?; | |
244 | Ok(()) | |
245 | } | |
246 | ||
247 | impl CinepakEncoder { | |
248 | fn new() -> Self { | |
249 | Self { | |
250 | stream: None, | |
251 | pkt: None, | |
252 | lastfrm: None, | |
253 | frmcount: 0, | |
254 | quality: 0, | |
255 | nstrips: 2, | |
256 | v1_entries: Vec::new(), | |
257 | v4_entries: Vec::new(), | |
258 | v1_cb: [YUVCode::default(); 256], | |
259 | v4_cb: [YUVCode::default(); 256], | |
260 | v1_cur_cb: [YUVCode::default(); 256], | |
261 | v4_cur_cb: [YUVCode::default(); 256], | |
262 | grayscale: false, | |
263 | rng: RNG::new(), | |
264 | v1_idx: Vec::new(), | |
265 | v4_idx: Vec::new(), | |
266 | masks: MaskWriter::new(), | |
267 | skip_dist: Vec::new(), | |
268 | } | |
269 | } | |
270 | fn read_strip(&mut self, in_frm: &NAVideoBuffer<u8>, start: usize, end: usize) { | |
271 | let ystride = in_frm.get_stride(0); | |
272 | let mut yoff = in_frm.get_offset(0) + start * ystride; | |
273 | let ustride = in_frm.get_stride(1); | |
274 | let mut uoff = in_frm.get_offset(1) + start / 2 * ustride; | |
275 | let vstride = in_frm.get_stride(2); | |
276 | let mut voff = in_frm.get_offset(2) + start / 2 * vstride; | |
277 | let (width, _) = in_frm.get_dimensions(0); | |
278 | let data = in_frm.get_data(); | |
279 | self.v1_entries.truncate(0); | |
280 | self.v4_entries.truncate(0); | |
281 | for _ in (start..end).step_by(4) { | |
282 | for x in (0..width).step_by(4) { | |
283 | let mut yblk = [0; 16]; | |
284 | let mut ublk = [128; 4]; | |
285 | let mut vblk = [128; 4]; | |
286 | for j in 0..4 { | |
287 | for i in 0..4 { | |
288 | yblk[i + j * 4] = data[yoff + x + i + j * ystride]; | |
289 | } | |
290 | } | |
291 | if !self.grayscale { | |
292 | for j in 0..2 { | |
293 | for i in 0..2 { | |
294 | ublk[i + j * 2] = data[uoff + x / 2 + i + j * ustride]; | |
295 | vblk[i + j * 2] = data[voff + x / 2 + i + j * vstride]; | |
296 | } | |
297 | } | |
298 | } | |
299 | self.v1_entries.push(YUVCode { | |
300 | y: [avg4(yblk[ 0], yblk[ 1], yblk[ 4], yblk[ 5]), | |
301 | avg4(yblk[ 2], yblk[ 3], yblk[ 6], yblk[ 7]), | |
302 | avg4(yblk[ 8], yblk[ 9], yblk[12], yblk[13]), | |
303 | avg4(yblk[10], yblk[11], yblk[14], yblk[15])], | |
304 | u: avg4(ublk[0], ublk[1], ublk[2], ublk[3]), | |
305 | v: avg4(vblk[0], vblk[1], vblk[2], vblk[3]), | |
306 | }); | |
307 | for i in 0..4 { | |
308 | let yidx = (i & 1) * 2 + (i & 2) * 4; | |
309 | self.v4_entries.push(YUVCode { | |
310 | y: [ yblk[yidx], yblk[yidx + 1], yblk[yidx + 4], yblk[yidx + 5] ], | |
311 | u: ublk[i], | |
312 | v: vblk[i], | |
313 | }); | |
314 | } | |
315 | } | |
316 | yoff += ystride * 4; | |
317 | uoff += ustride * 2; | |
318 | voff += vstride * 2; | |
319 | } | |
320 | } | |
321 | fn find_nearest(codebook: &[YUVCode; 256], code: YUVCode) -> (u8, u32) { | |
322 | let mut min_dist = std::u32::MAX; | |
323 | let mut idx = 0; | |
324 | for (i, cw) in codebook.iter().enumerate() { | |
325 | let dist = cw.dist(code); | |
326 | if dist < min_dist { | |
327 | min_dist = dist; | |
328 | idx = i; | |
329 | if dist == 0 { | |
330 | break; | |
331 | } | |
332 | } | |
333 | } | |
334 | (idx as u8, min_dist) | |
335 | } | |
336 | fn can_update_cb(new_cb: &[YUVCode; 256], old_cb: &[YUVCode; 256], cb_size: usize) -> bool { | |
337 | let mut skip_count = 0; | |
338 | for (new, old) in new_cb.iter().zip(old_cb.iter()) { | |
339 | if new == old { | |
340 | skip_count += 1; | |
341 | } | |
342 | } | |
343 | let full_size = cb_size * 256; | |
344 | let upd_size = cb_size * (256 - skip_count) + 64; | |
345 | upd_size < full_size | |
346 | } | |
347 | fn write_cb(bw: &mut ByteWriter, mut id: u8, new_cb: &[YUVCode; 256], old_cb: &[YUVCode; 256], grayscale: bool, update: bool) -> EncoderResult<()> { | |
348 | if grayscale { | |
349 | id |= 4; | |
350 | } | |
351 | if update { | |
352 | id |= 1; | |
353 | } | |
354 | bw.write_byte(id)?; | |
355 | bw.write_u24be(0)?; | |
356 | let chunk_pos = bw.tell(); | |
357 | if !update { | |
358 | for entry in new_cb.iter() { | |
359 | bw.write_buf(&entry.y)?; | |
360 | if !grayscale { | |
361 | bw.write_byte(entry.u ^ 0x80)?; | |
362 | bw.write_byte(entry.v ^ 0x80)?; | |
363 | } | |
364 | } | |
365 | } else { | |
366 | let mut end = 256; | |
367 | for (i, (ncw, ocw)) in new_cb.iter().rev().zip(old_cb.iter().rev()).enumerate() { | |
368 | if ncw == ocw { | |
369 | end = i; | |
370 | } else { | |
371 | break; | |
372 | } | |
373 | } | |
374 | for i in (0..end).step_by(32) { | |
375 | let mut mask = 0; | |
376 | for j in 0..32 { | |
377 | mask <<= 1; | |
378 | if new_cb[i + j] != old_cb[i + j] { | |
379 | mask |= 1; | |
380 | } | |
381 | } | |
382 | bw.write_u32be(mask)?; | |
383 | for j in 0..32 { | |
384 | if new_cb[i + j] == old_cb[i + j] { continue; } | |
385 | bw.write_buf(&new_cb[i + j].y)?; | |
386 | if !grayscale { | |
387 | bw.write_byte(new_cb[i + j].u ^ 0x80)?; | |
388 | bw.write_byte(new_cb[i + j].v ^ 0x80)?; | |
389 | } | |
390 | } | |
391 | } | |
392 | } | |
393 | patch_size(bw, chunk_pos)?; | |
394 | Ok(()) | |
395 | } | |
396 | fn render_stripe(&mut self, intra: bool, start: usize, end: usize) { | |
397 | if let Some(ref mut dst_frm) = self.lastfrm { | |
398 | let ystride = dst_frm.get_stride(0); | |
399 | let mut yoff = dst_frm.get_offset(0) + start * ystride; | |
400 | let ustride = dst_frm.get_stride(1); | |
401 | let mut uoff = dst_frm.get_offset(1) + start / 2 * ustride; | |
402 | let vstride = dst_frm.get_stride(2); | |
403 | let mut voff = dst_frm.get_offset(2) + start / 2 * vstride; | |
404 | let (width, _) = dst_frm.get_dimensions(0); | |
405 | let data = dst_frm.get_data_mut().unwrap(); | |
406 | let mut miter = self.masks.masks.iter(); | |
407 | let mut v1_iter = self.v1_idx.iter(); | |
408 | let mut v4_iter = self.v4_idx.iter(); | |
409 | let mut cur_mask = 0; | |
410 | let mut cur_bit = 0; | |
411 | for _ in (start..end).step_by(4) { | |
412 | for x in (0..width).step_by(4) { | |
413 | if cur_bit == 0 { | |
414 | if !intra || self.v1_idx.len() > 0 { | |
415 | cur_mask = *miter.next().unwrap(); | |
416 | } else { | |
417 | cur_mask = 0xFFFFFFFF; | |
418 | } | |
419 | cur_bit = 1 << 31; | |
420 | } | |
421 | if !intra { | |
422 | if (cur_mask & cur_bit) == 0 { | |
423 | cur_bit >>= 1; | |
424 | continue; | |
425 | } | |
426 | cur_bit >>= 1; | |
427 | if cur_bit == 0 { | |
428 | cur_mask = *miter.next().unwrap(); | |
429 | cur_bit = 1 << 31; | |
430 | } | |
431 | } | |
432 | if (cur_mask & cur_bit) == 0 { | |
433 | let idx = *v1_iter.next().unwrap() as usize; | |
434 | let cb = &self.v1_cur_cb[idx]; | |
435 | ||
436 | let mut coff = yoff + x; | |
437 | data[coff] = cb.y[0]; data[coff + 1] = cb.y[0]; | |
438 | data[coff + 2] = cb.y[1]; data[coff + 3] = cb.y[1]; | |
439 | coff += ystride; | |
440 | data[coff] = cb.y[0]; data[coff + 1] = cb.y[0]; | |
441 | data[coff + 2] = cb.y[1]; data[coff + 3] = cb.y[1]; | |
442 | coff += ystride; | |
443 | data[coff] = cb.y[2]; data[coff + 1] = cb.y[2]; | |
444 | data[coff + 2] = cb.y[3]; data[coff + 3] = cb.y[3]; | |
445 | coff += ystride; | |
446 | data[coff] = cb.y[2]; data[coff + 1] = cb.y[2]; | |
447 | data[coff + 2] = cb.y[3]; data[coff + 3] = cb.y[3]; | |
448 | ||
449 | if !self.grayscale { | |
450 | let mut coff = uoff + x / 2; | |
451 | data[coff] = cb.u; data[coff + 1] = cb.u; | |
452 | coff += ustride; | |
453 | data[coff] = cb.u; data[coff + 1] = cb.u; | |
454 | ||
455 | let mut coff = voff + x / 2; | |
456 | data[coff] = cb.v; data[coff + 1] = cb.v; | |
457 | coff += vstride; | |
458 | data[coff] = cb.v; data[coff + 1] = cb.v; | |
459 | } | |
460 | } else { | |
461 | let idx0 = *v4_iter.next().unwrap() as usize; | |
462 | let cb0 = &self.v4_cur_cb[idx0]; | |
463 | let idx1 = *v4_iter.next().unwrap() as usize; | |
464 | let cb1 = &self.v4_cur_cb[idx1]; | |
465 | let idx2 = *v4_iter.next().unwrap() as usize; | |
466 | let cb2 = &self.v4_cur_cb[idx2]; | |
467 | let idx3 = *v4_iter.next().unwrap() as usize; | |
468 | let cb3 = &self.v4_cur_cb[idx3]; | |
469 | ||
470 | let mut coff = yoff + x; | |
471 | data[coff] = cb0.y[0]; data[coff + 1] = cb0.y[1]; | |
472 | data[coff + 2] = cb1.y[0]; data[coff + 3] = cb1.y[1]; | |
473 | coff += ystride; | |
474 | data[coff] = cb0.y[2]; data[coff + 1] = cb0.y[3]; | |
475 | data[coff + 2] = cb1.y[2]; data[coff + 3] = cb1.y[3]; | |
476 | coff += ystride; | |
477 | data[coff] = cb2.y[0]; data[coff + 1] = cb2.y[1]; | |
478 | data[coff + 2] = cb3.y[0]; data[coff + 3] = cb3.y[1]; | |
479 | coff += ystride; | |
480 | data[coff] = cb2.y[2]; data[coff + 1] = cb2.y[3]; | |
481 | data[coff + 2] = cb3.y[2]; data[coff + 3] = cb3.y[3]; | |
482 | ||
483 | if !self.grayscale { | |
484 | let mut coff = uoff + x / 2; | |
485 | data[coff] = cb0.u; data[coff + 1] = cb1.u; | |
486 | coff += ustride; | |
487 | data[coff] = cb2.u; data[coff + 1] = cb3.u; | |
488 | ||
489 | let mut coff = voff + x / 2; | |
490 | data[coff] = cb0.v; data[coff + 1] = cb1.v; | |
491 | coff += vstride; | |
492 | data[coff] = cb2.v; data[coff + 1] = cb3.v; | |
493 | } | |
494 | } | |
495 | cur_bit >>= 1; | |
496 | } | |
497 | yoff += ystride * 4; | |
498 | uoff += ustride * 2; | |
499 | voff += vstride * 2; | |
500 | } | |
501 | } else { | |
502 | unreachable!(); | |
503 | } | |
504 | } | |
505 | fn calc_skip_dist(&mut self, in_frm: &NAVideoBuffer<u8>, start: usize, end: usize) { | |
506 | self.skip_dist.truncate(0); | |
507 | if let Some(ref ref_frm) = self.lastfrm { | |
508 | let rystride = ref_frm.get_stride(0); | |
509 | let mut ryoff = ref_frm.get_offset(0) + start * rystride; | |
510 | let rustride = ref_frm.get_stride(1); | |
511 | let mut ruoff = ref_frm.get_offset(1) + start / 2 * rustride; | |
512 | let rvstride = ref_frm.get_stride(2); | |
513 | let mut rvoff = ref_frm.get_offset(2) + start / 2 * rvstride; | |
514 | let (width, _) = ref_frm.get_dimensions(0); | |
515 | let rdata = ref_frm.get_data(); | |
516 | ||
517 | let iystride = in_frm.get_stride(0); | |
518 | let mut iyoff = in_frm.get_offset(0) + start * iystride; | |
519 | let iustride = in_frm.get_stride(1); | |
520 | let mut iuoff = in_frm.get_offset(1) + start / 2 * iustride; | |
521 | let ivstride = in_frm.get_stride(2); | |
522 | let mut ivoff = in_frm.get_offset(2) + start / 2 * ivstride; | |
523 | let idata = in_frm.get_data(); | |
524 | ||
525 | for _ in (start..end).step_by(4) { | |
526 | for x in (0..width).step_by(4) { | |
527 | let mut dist = 0; | |
528 | let mut roff = ryoff + x; | |
529 | let mut ioff = iyoff + x; | |
530 | for _ in 0..4 { | |
531 | for i in 0..4 { | |
532 | let d = i32::from(rdata[roff + i]) - i32::from(idata[ioff + i]); | |
533 | dist += d * d; | |
534 | } | |
535 | roff += rystride; | |
536 | ioff += iystride; | |
537 | } | |
538 | if !self.grayscale { | |
539 | let mut roff = ruoff + x / 2; | |
540 | let mut ioff = iuoff + x / 2; | |
541 | let ud = i32::from(rdata[roff]) - i32::from(idata[ioff]); | |
542 | dist += ud * ud; | |
543 | let ud = i32::from(rdata[roff + 1]) - i32::from(idata[ioff + 1]); | |
544 | dist += ud * ud; | |
545 | roff += rustride; ioff += iustride; | |
546 | let ud = i32::from(rdata[roff]) - i32::from(idata[ioff]); | |
547 | dist += ud * ud; | |
548 | let ud = i32::from(rdata[roff + 1]) - i32::from(idata[ioff + 1]); | |
549 | dist += ud * ud; | |
550 | ||
551 | let mut roff = rvoff + x / 2; | |
552 | let mut ioff = ivoff + x / 2; | |
553 | let vd = i32::from(rdata[roff]) - i32::from(idata[ioff]); | |
554 | dist += vd * vd; | |
555 | let vd = i32::from(rdata[roff + 1]) - i32::from(idata[ioff + 1]); | |
556 | dist += vd * vd; | |
557 | roff += rvstride; ioff += ivstride; | |
558 | let vd = i32::from(rdata[roff]) - i32::from(idata[ioff]); | |
559 | dist += vd * vd; | |
560 | let vd = i32::from(rdata[roff + 1]) - i32::from(idata[ioff + 1]); | |
561 | dist += vd * vd; | |
562 | } | |
563 | self.skip_dist.push(dist as u32); | |
564 | } | |
565 | ||
566 | iyoff += iystride * 4; | |
567 | iuoff += iustride * 2; | |
568 | ivoff += ivstride * 2; | |
569 | ryoff += rystride * 4; | |
570 | ruoff += rustride * 2; | |
571 | rvoff += rvstride * 2; | |
572 | } | |
573 | } else { | |
574 | unreachable!(); | |
575 | } | |
576 | } | |
577 | fn encode_intra(&mut self, bw: &mut ByteWriter, in_frm: &NAVideoBuffer<u8>) -> EncoderResult<bool> { | |
578 | let (width, height) = in_frm.get_dimensions(0); | |
579 | let mut strip_h = (height / self.nstrips + 3) & !3; | |
580 | if strip_h == 0 { | |
581 | self.nstrips = 1; | |
582 | strip_h = height; | |
583 | } | |
584 | let mut start_line = 0; | |
585 | let mut end_line = strip_h; | |
586 | ||
587 | bw.write_byte(0)?; // intra flag | |
588 | bw.write_u24be(0)?; // frame size | |
589 | let frame_data_pos = bw.tell(); | |
590 | bw.write_u16be(width as u16)?; | |
591 | bw.write_u16be(height as u16)?; | |
592 | bw.write_u16be(self.nstrips as u16)?; | |
593 | ||
594 | for entry in self.v1_cb.iter_mut() { | |
595 | self.rng.fill_entry(entry); | |
596 | } | |
597 | for entry in self.v4_cb.iter_mut() { | |
598 | self.rng.fill_entry(entry); | |
599 | } | |
600 | while start_line < height { | |
601 | self.read_strip(in_frm, start_line, end_line); | |
602 | ||
603 | // let mut elbg_v1: ELBG<YUVCode, YUVCodeSum> = ELBG::new(&self.v1_cb); | |
604 | // let mut elbg_v4: ELBG<YUVCode, YUVCodeSum> = ELBG::new(&self.v4_cb); | |
605 | // elbg_v1.quantise(&self.v1_entries, &mut self.v1_cur_cb); | |
606 | // elbg_v4.quantise(&self.v4_entries, &mut self.v4_cur_cb); | |
607 | quantise_median_cut::<YUVCode, YUVCodeSum>(&self.v1_entries, &mut self.v1_cur_cb); | |
608 | quantise_median_cut::<YUVCode, YUVCodeSum>(&self.v4_entries, &mut self.v4_cur_cb); | |
609 | if self.grayscale { | |
610 | for cw in self.v1_cur_cb.iter_mut() { | |
611 | cw.u = 128; | |
612 | cw.v = 128; | |
613 | } | |
614 | for cw in self.v4_cur_cb.iter_mut() { | |
615 | cw.u = 128; | |
616 | cw.v = 128; | |
617 | } | |
618 | } | |
619 | ||
620 | self.v1_idx.truncate(0); | |
621 | self.v4_idx.truncate(0); | |
622 | self.masks.reset(); | |
623 | ||
624 | for (v1_entry, v4_entries) in self.v1_entries.iter().zip(self.v4_entries.chunks(4)) { | |
625 | let (v1_idx, v1_dist) = Self::find_nearest(&self.v1_cur_cb, *v1_entry); | |
626 | if v1_dist == 0 { | |
627 | self.masks.put_v1(); | |
628 | self.v1_idx.push(v1_idx); | |
629 | continue; | |
630 | } | |
631 | let (v40_idx, v40_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[0]); | |
632 | let (v41_idx, v41_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[1]); | |
633 | let (v42_idx, v42_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[2]); | |
634 | let (v43_idx, v43_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[3]); | |
635 | if v40_dist + v41_dist + v42_dist + v43_dist > v1_dist { | |
636 | self.masks.put_v4(); | |
637 | self.v4_idx.push(v40_idx); | |
638 | self.v4_idx.push(v41_idx); | |
639 | self.v4_idx.push(v42_idx); | |
640 | self.v4_idx.push(v43_idx); | |
641 | } else { | |
642 | self.masks.put_v1(); | |
643 | self.v1_idx.push(v1_idx); | |
644 | } | |
645 | } | |
646 | self.masks.end(); | |
647 | ||
648 | let mut is_intra_strip = start_line == 0; | |
649 | let (upd_v1, upd_v4) = if !is_intra_strip { | |
650 | let cb_size = if self.grayscale { 4 } else { 6 }; | |
651 | (Self::can_update_cb(&self.v1_cur_cb, &self.v1_cb, cb_size), | |
652 | Self::can_update_cb(&self.v4_cur_cb, &self.v4_cb, cb_size)) | |
653 | } else { | |
654 | (false, false) | |
655 | }; | |
656 | if !is_intra_strip && !upd_v1 && !upd_v4 { | |
657 | is_intra_strip = true; | |
658 | } | |
659 | bw.write_byte(if is_intra_strip { 0x10 } else { 0x11 })?; | |
660 | bw.write_u24be(0)?; // strip size | |
661 | let strip_data_pos = bw.tell(); | |
662 | bw.write_u16be(0)?; // yoff | |
663 | bw.write_u16be(0)?; // xoff | |
664 | bw.write_u16be((end_line - start_line) as u16)?; | |
665 | bw.write_u16be(width as u16)?; | |
666 | ||
667 | Self::write_cb(bw, 0x20, &self.v4_cur_cb, &self.v4_cb, self.grayscale, upd_v4)?; | |
668 | Self::write_cb(bw, 0x22, &self.v1_cur_cb, &self.v1_cb, self.grayscale, upd_v1)?; | |
669 | ||
670 | self.render_stripe(true, start_line, end_line); | |
671 | ||
672 | if self.v1_idx.len() == 0 { | |
673 | bw.write_byte(0x32)?; | |
674 | bw.write_u24be((self.v4_idx.len() + 4) as u32)?; | |
675 | bw.write_buf(self.v4_idx.as_slice())?; | |
676 | } else { | |
677 | bw.write_byte(0x30)?; | |
678 | bw.write_u24be(0)?; | |
679 | let chunk_pos = bw.tell(); | |
680 | let mut v1_pos = 0; | |
681 | let mut v4_pos = 0; | |
682 | for _ in 0..32 { | |
683 | self.v1_idx.push(0); | |
684 | self.v4_idx.push(0); | |
685 | self.v4_idx.push(0); | |
686 | self.v4_idx.push(0); | |
687 | self.v4_idx.push(0); | |
688 | } | |
689 | for mask in self.masks.masks.iter() { | |
690 | bw.write_u32be(*mask)?; | |
691 | for j in (0..32).rev() { | |
692 | if (mask & (1 << j)) == 0 { | |
693 | bw.write_byte(self.v1_idx[v1_pos])?; | |
694 | v1_pos += 1; | |
695 | } else { | |
696 | bw.write_byte(self.v4_idx[v4_pos])?; | |
697 | bw.write_byte(self.v4_idx[v4_pos + 1])?; | |
698 | bw.write_byte(self.v4_idx[v4_pos + 2])?; | |
699 | bw.write_byte(self.v4_idx[v4_pos + 3])?; | |
700 | v4_pos += 4; | |
701 | } | |
702 | } | |
703 | } | |
704 | patch_size(bw, chunk_pos)?; | |
705 | } | |
706 | ||
707 | patch_size(bw, strip_data_pos)?; | |
708 | ||
709 | self.v1_cb.copy_from_slice(&self.v1_cur_cb); | |
710 | self.v4_cb.copy_from_slice(&self.v4_cur_cb); | |
711 | start_line = end_line; | |
712 | end_line = (end_line + strip_h).min(height); | |
713 | } | |
714 | patch_size(bw, frame_data_pos)?; | |
715 | Ok(true) | |
716 | } | |
717 | fn encode_inter(&mut self, bw: &mut ByteWriter, in_frm: &NAVideoBuffer<u8>) -> EncoderResult<bool> { | |
718 | let (width, height) = in_frm.get_dimensions(0); | |
719 | let mut strip_h = (height / self.nstrips + 3) & !3; | |
720 | if strip_h == 0 { | |
721 | self.nstrips = 1; | |
722 | strip_h = height; | |
723 | } | |
724 | let mut start_line = 0; | |
725 | let mut end_line = strip_h; | |
726 | ||
727 | bw.write_byte(1)?; // intra flag | |
728 | bw.write_u24be(0)?; // frame size | |
729 | let frame_data_pos = bw.tell(); | |
730 | bw.write_u16be(width as u16)?; | |
731 | bw.write_u16be(height as u16)?; | |
732 | bw.write_u16be(self.nstrips as u16)?; | |
733 | ||
734 | while start_line < height { | |
735 | self.read_strip(in_frm, start_line, end_line); | |
736 | self.calc_skip_dist(in_frm, start_line, end_line); | |
737 | ||
738 | // let mut elbg_v1: ELBG<YUVCode, YUVCodeSum> = ELBG::new(&self.v1_cb); | |
739 | // let mut elbg_v4: ELBG<YUVCode, YUVCodeSum> = ELBG::new(&self.v4_cb); | |
740 | // elbg_v1.quantise(&self.v1_entries, &mut self.v1_cur_cb); | |
741 | // elbg_v4.quantise(&self.v4_entries, &mut self.v4_cur_cb); | |
742 | quantise_median_cut::<YUVCode, YUVCodeSum>(&self.v1_entries, &mut self.v1_cur_cb); | |
743 | quantise_median_cut::<YUVCode, YUVCodeSum>(&self.v4_entries, &mut self.v4_cur_cb); | |
744 | if self.grayscale { | |
745 | for cw in self.v1_cur_cb.iter_mut() { | |
746 | cw.u = 128; | |
747 | cw.v = 128; | |
748 | } | |
749 | for cw in self.v4_cur_cb.iter_mut() { | |
750 | cw.u = 128; | |
751 | cw.v = 128; | |
752 | } | |
753 | } | |
754 | ||
755 | self.v1_idx.truncate(0); | |
756 | self.v4_idx.truncate(0); | |
757 | self.masks.reset(); | |
758 | ||
759 | let mut skip_iter = self.skip_dist.iter(); | |
760 | for (v1_entry, v4_entries) in self.v1_entries.iter().zip(self.v4_entries.chunks(4)) { | |
761 | let skip_dist = *skip_iter.next().unwrap(); | |
762 | if skip_dist == 0 { | |
763 | self.masks.put_inter(true); | |
764 | continue; | |
765 | } | |
766 | let (v1_idx, v1_dist) = Self::find_nearest(&self.v1_cur_cb, *v1_entry); | |
767 | if skip_dist < v1_dist { | |
768 | self.masks.put_inter(true); | |
769 | continue; | |
770 | } else { | |
771 | self.masks.put_inter(false); | |
772 | } | |
773 | if v1_dist == 0 { | |
774 | self.masks.put_v1(); | |
775 | self.v1_idx.push(v1_idx); | |
776 | continue; | |
777 | } | |
778 | let (v40_idx, v40_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[0]); | |
779 | let (v41_idx, v41_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[1]); | |
780 | let (v42_idx, v42_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[2]); | |
781 | let (v43_idx, v43_dist) = Self::find_nearest(&self.v4_cur_cb, v4_entries[3]); | |
782 | if v40_dist + v41_dist + v42_dist + v43_dist > v1_dist { | |
783 | self.masks.put_v4(); | |
784 | self.v4_idx.push(v40_idx); | |
785 | self.v4_idx.push(v41_idx); | |
786 | self.v4_idx.push(v42_idx); | |
787 | self.v4_idx.push(v43_idx); | |
788 | } else { | |
789 | self.masks.put_v1(); | |
790 | self.v1_idx.push(v1_idx); | |
791 | } | |
792 | } | |
793 | self.masks.end(); | |
794 | ||
795 | let (upd_v1, upd_v4) = { | |
796 | let cb_size = if self.grayscale { 4 } else { 6 }; | |
797 | (Self::can_update_cb(&self.v1_cur_cb, &self.v1_cb, cb_size), | |
798 | Self::can_update_cb(&self.v4_cur_cb, &self.v4_cb, cb_size)) | |
799 | }; | |
800 | bw.write_byte(0x11)?; | |
801 | bw.write_u24be(0)?; // strip size | |
802 | let strip_data_pos = bw.tell(); | |
803 | bw.write_u16be(0)?; // yoff | |
804 | bw.write_u16be(0)?; // xoff | |
805 | bw.write_u16be((end_line - start_line) as u16)?; | |
806 | bw.write_u16be(width as u16)?; | |
807 | ||
808 | Self::write_cb(bw, 0x20, &self.v4_cur_cb, &self.v4_cb, self.grayscale, upd_v4)?; | |
809 | Self::write_cb(bw, 0x22, &self.v1_cur_cb, &self.v1_cb, self.grayscale, upd_v1)?; | |
810 | ||
811 | self.render_stripe(false, start_line, end_line); | |
812 | ||
813 | bw.write_byte(0x31)?; | |
814 | bw.write_u24be(0)?; | |
815 | let chunk_pos = bw.tell(); | |
816 | let mut v1_pos = 0; | |
817 | let mut v4_pos = 0; | |
818 | for _ in 0..32 { | |
819 | self.v1_idx.push(0); | |
820 | self.v4_idx.push(0); | |
821 | self.v4_idx.push(0); | |
822 | self.v4_idx.push(0); | |
823 | self.v4_idx.push(0); | |
824 | } | |
825 | let mut skip = true; | |
826 | for mask in self.masks.masks.iter() { | |
827 | bw.write_u32be(*mask)?; | |
828 | if *mask == 0 { continue; } | |
829 | let mut bit = 1 << 31; | |
830 | while bit > 0 { | |
831 | if skip { | |
832 | skip = (mask & bit) == 0; | |
833 | bit >>= 1; | |
834 | } else { | |
835 | if (mask & bit) == 0 { | |
836 | bw.write_byte(self.v1_idx[v1_pos])?; | |
837 | v1_pos += 1; | |
838 | } else { | |
839 | bw.write_byte(self.v4_idx[v4_pos])?; | |
840 | bw.write_byte(self.v4_idx[v4_pos + 1])?; | |
841 | bw.write_byte(self.v4_idx[v4_pos + 2])?; | |
842 | bw.write_byte(self.v4_idx[v4_pos + 3])?; | |
843 | v4_pos += 4; | |
844 | } | |
845 | bit >>= 1; | |
846 | skip = true; | |
847 | } | |
848 | } | |
849 | } | |
850 | patch_size(bw, chunk_pos)?; | |
851 | ||
852 | patch_size(bw, strip_data_pos)?; | |
853 | ||
854 | self.v1_cb.copy_from_slice(&self.v1_cur_cb); | |
855 | self.v4_cb.copy_from_slice(&self.v4_cur_cb); | |
856 | start_line = end_line; | |
857 | end_line = (end_line + strip_h).min(height); | |
858 | } | |
859 | patch_size(bw, frame_data_pos)?; | |
860 | Ok(true) | |
861 | } | |
862 | } | |
863 | ||
864 | impl NAEncoder for CinepakEncoder { | |
865 | fn negotiate_format(&self, encinfo: &EncodeParameters) -> EncoderResult<EncodeParameters> { | |
866 | match encinfo.format { | |
867 | NACodecTypeInfo::None => { | |
868 | let mut ofmt = EncodeParameters::default(); | |
869 | ofmt.format = NACodecTypeInfo::Video(NAVideoInfo::new(0, 0, true, YUV420_FORMAT)); | |
870 | Ok(ofmt) | |
871 | }, | |
872 | NACodecTypeInfo::Audio(_) => return Err(EncoderError::FormatError), | |
873 | NACodecTypeInfo::Video(vinfo) => { | |
874 | let pix_fmt = if vinfo.format == GRAY_FORMAT { GRAY_FORMAT } else { YUV420_FORMAT }; | |
875 | let outinfo = NAVideoInfo::new((vinfo.width + 3) & !3, (vinfo.height + 3) & !3, true, pix_fmt); | |
d722ffe9 | 876 | let mut ofmt = *encinfo; |
a178c22c KS |
877 | ofmt.format = NACodecTypeInfo::Video(outinfo); |
878 | Ok(ofmt) | |
879 | } | |
880 | } | |
881 | } | |
882 | fn init(&mut self, stream_id: u32, encinfo: EncodeParameters) -> EncoderResult<NAStreamRef> { | |
883 | match encinfo.format { | |
884 | NACodecTypeInfo::None => Err(EncoderError::FormatError), | |
885 | NACodecTypeInfo::Audio(_) => Err(EncoderError::FormatError), | |
886 | NACodecTypeInfo::Video(vinfo) => { | |
887 | if vinfo.format != YUV420_FORMAT && vinfo.format != GRAY_FORMAT { | |
888 | return Err(EncoderError::FormatError); | |
889 | } | |
890 | if ((vinfo.width | vinfo.height) & 3) != 0 { | |
891 | return Err(EncoderError::FormatError); | |
892 | } | |
893 | if (vinfo.width | vinfo.height) >= (1 << 16) { | |
894 | return Err(EncoderError::FormatError); | |
895 | } | |
896 | ||
897 | let out_info = NAVideoInfo::new(vinfo.width, vinfo.height, false, vinfo.format); | |
898 | let info = NACodecInfo::new("cinepak", NACodecTypeInfo::Video(out_info.clone()), None); | |
899 | let stream = NAStream::new(StreamType::Video, stream_id, info, encinfo.tb_num, encinfo.tb_den).into_ref(); | |
900 | ||
901 | self.stream = Some(stream.clone()); | |
902 | self.quality = encinfo.quality; | |
903 | self.grayscale = vinfo.format != YUV420_FORMAT; | |
904 | let num_blocks = vinfo.width / 2 * vinfo.height / 2; | |
905 | self.v1_entries = Vec::with_capacity(num_blocks); | |
906 | self.v4_entries = Vec::with_capacity(num_blocks * 4); | |
907 | self.v1_idx = Vec::with_capacity(num_blocks); | |
908 | self.v4_idx = Vec::with_capacity(num_blocks * 4); | |
909 | self.skip_dist = Vec::with_capacity(vinfo.width / 4 * vinfo.height / 4); | |
910 | ||
911 | let buf = alloc_video_buffer(out_info, 2)?; | |
912 | self.lastfrm = Some(buf.get_vbuf().unwrap()); | |
913 | ||
914 | Ok(stream) | |
915 | }, | |
916 | } | |
917 | } | |
918 | fn encode(&mut self, frm: &NAFrame) -> EncoderResult<()> { | |
919 | let buf = frm.get_buffer(); | |
920 | if let Some(ref vbuf) = buf.get_vbuf() { | |
921 | let mut dbuf = Vec::with_capacity(4); | |
922 | let mut gw = GrowableMemoryWriter::new_write(&mut dbuf); | |
923 | let mut bw = ByteWriter::new(&mut gw); | |
924 | let is_intra = if self.frmcount == 0 { | |
925 | self.encode_intra(&mut bw, vbuf)? | |
926 | } else { | |
927 | self.encode_inter(&mut bw, vbuf)? | |
928 | }; | |
929 | self.pkt = Some(NAPacket::new(self.stream.clone().unwrap(), frm.ts, is_intra, dbuf)); | |
930 | self.frmcount += 1; | |
931 | if self.frmcount == 25 { | |
932 | self.frmcount = 0; | |
933 | } | |
934 | Ok(()) | |
935 | } else { | |
936 | Err(EncoderError::InvalidParameters) | |
937 | } | |
938 | } | |
939 | fn get_packet(&mut self) -> EncoderResult<Option<NAPacket>> { | |
940 | let mut npkt = None; | |
941 | std::mem::swap(&mut self.pkt, &mut npkt); | |
942 | Ok(npkt) | |
943 | } | |
944 | fn flush(&mut self) -> EncoderResult<()> { | |
945 | self.frmcount = 0; | |
946 | Ok(()) | |
947 | } | |
948 | } | |
949 | ||
950 | impl NAOptionHandler for CinepakEncoder { | |
951 | fn get_supported_options(&self) -> &[NAOptionDefinition] { &[] } | |
952 | fn set_options(&mut self, _options: &[NAOption]) { } | |
953 | fn query_option_value(&self, _name: &str) -> Option<NAValue> { None } | |
954 | } | |
955 | ||
956 | pub fn get_encoder() -> Box<dyn NAEncoder + Send> { | |
957 | Box::new(CinepakEncoder::new()) | |
958 | } | |
959 | ||
960 | #[cfg(test)] | |
961 | mod test { | |
962 | use nihav_core::codecs::*; | |
963 | use nihav_core::demuxers::*; | |
964 | use nihav_core::muxers::*; | |
965 | use crate::*; | |
966 | use nihav_codec_support::test::enc_video::*; | |
967 | ||
968 | #[test] | |
969 | fn test_cinepak_encoder() { | |
970 | let mut dmx_reg = RegisteredDemuxers::new(); | |
971 | generic_register_all_demuxers(&mut dmx_reg); | |
972 | let mut dec_reg = RegisteredDecoders::new(); | |
973 | generic_register_all_codecs(&mut dec_reg); | |
974 | let mut mux_reg = RegisteredMuxers::new(); | |
975 | generic_register_all_muxers(&mut mux_reg); | |
976 | let mut enc_reg = RegisteredEncoders::new(); | |
977 | generic_register_all_encoders(&mut enc_reg); | |
978 | ||
979 | let dec_config = DecoderTestParams { | |
980 | demuxer: "avi", | |
981 | in_name: "assets/Misc/TalkingHead_352x288.avi", | |
982 | stream_type: StreamType::Video, | |
983 | limit: Some(2), | |
984 | dmx_reg, dec_reg, | |
985 | }; | |
986 | let enc_config = EncoderTestParams { | |
987 | muxer: "avi", | |
988 | enc_name: "cinepak", | |
989 | out_name: "cinepak.avi", | |
990 | mux_reg, enc_reg, | |
991 | }; | |
992 | let dst_vinfo = NAVideoInfo { | |
993 | width: 0, | |
994 | height: 0, | |
995 | format: YUV420_FORMAT, | |
996 | flipped: true, | |
997 | }; | |
998 | let enc_params = EncodeParameters { | |
999 | format: NACodecTypeInfo::Video(dst_vinfo), | |
1000 | quality: 0, | |
1001 | bitrate: 0, | |
1002 | tb_num: 0, | |
1003 | tb_den: 0, | |
1004 | flags: 0, | |
1005 | }; | |
1006 | test_encoding_to_file(&dec_config, &enc_config, enc_params); | |
1007 | } | |
1008 | } |