oxedyne/fe2o3/fe2o3_graphics/src/hevc/mod.rs
54.2 KiB, 280 runs
created by r1870400018:20461, which is this file's identity for as long as the history lasts, whatever it is later renamed to
download · who wrote it · its history
| 1 | //! An HEVC decoder, for the still pictures inside a HEIC file. |
| 2 | //! |
| 3 | //! HEVC (ITU-T H.265) is the codec a phone's photograph is coded in once it stops being JPEG, and |
| 4 | //! there is no way to read one without decoding it. This module is that decoder. It is built for |
| 5 | //! **intra** coding only -- a still picture refers to nothing but itself, so everything about |
| 6 | //! motion, reference pictures and prediction between frames is absent by construction rather than |
| 7 | //! unimplemented. |
| 8 | //! |
| 9 | //! # What is here |
| 10 | //! |
| 11 | //! The whole of it, for the intra case. The bitstream side: splitting a stream into NAL units, |
| 12 | //! undoing the emulation-prevention bytes, and reading the sequence and picture parameter sets that |
| 13 | //! say how big the picture is and how it is cut up. That is the part every later stage is written |
| 14 | //! against, and the part that can be checked before any pixel exists: the size a sequence parameter |
| 15 | //! set codes must agree with the size the HEIF container's `ispe` property declares, and those two |
| 16 | //! numbers are written into the file by different parts of an encoder. |
| 17 | //! |
| 18 | //! Then the slice segment header, including the entry points that say where each row of the picture |
| 19 | //! begins; the CABAC arithmetic decoder; the context variables each syntax element uses; the coding |
| 20 | //! quadtree; residual coding; dequantisation and the inverse transforms; all thirty-five intra |
| 21 | //! prediction modes; deblocking and the sample adaptive offset; and the conversion out of 4:2:0 |
| 22 | //! into red, green and blue. [`picture`] is the entry point and runs the lot. |
| 23 | //! |
| 24 | //! The arithmetic decoder is the last piece that can be held to a standard before a picture comes |
| 25 | //! out, and it is held to two: every context starts in a state the probability tables actually |
| 26 | //! have -- all 256 initialisation values against every quantisation parameter a slice may carry -- |
| 27 | //! and the coding interval is between 256 and 510 after every bin, whatever is fed in. A |
| 28 | //! renormalisation one shift short satisfies neither and decodes plausible rubbish rather than |
| 29 | //! failing, which is the kind of fault that otherwise survives until a photograph comes out |
| 30 | //! wrong. |
| 31 | //! |
| 32 | //! Everything after it is held to another decoder instead, because by then there is a picture to |
| 33 | //! compare: `tests/hevc_tiles.rs` puts every brightness and colour sample beside what FFmpeg makes |
| 34 | //! of the same file, with both loop filters running at both ends. |
| 35 | //! |
| 36 | //! # What the pictures in one real library actually are |
| 37 | //! |
| 38 | //! Every sequence parameter set in 359 HEIC photographs out of a family library was read, and they |
| 39 | //! are uniform: **8-bit 4:2:0, coding tree blocks of 32, the sample adaptive offset on, no PCM, no |
| 40 | //! scaling lists, one tile**. The tiles are 512 by 512 in 350 of them, 1024 by 1024 in eight, and |
| 41 | //! the one photograph not stored as a grid is 720 by 720. |
| 42 | //! |
| 43 | //! One of those measurements was a surprise and it changes the shape of the decoder: **every one of |
| 44 | //! them is coded in wavefronts** (`entropy_coding_sync_enabled_flag`). The arithmetic decoder is |
| 45 | //! therefore reset at the start of every row of coding tree blocks, from the state saved after the |
| 46 | //! second block of the row above, and the slice header carries a byte offset for each row. That is |
| 47 | //! not an exotic case to be refused; it is the case. All 359 slice headers read, and in each the |
| 48 | //! number of rows the header names agrees with the number the sequence parameter set implies -- |
| 49 | //! sixteen for a 512-pixel tile at 32, twenty-three for the 720-pixel picture -- which is the |
| 50 | //! check that caught the first reading, where the flag was mistaken for something rare and every |
| 51 | //! photograph in the corpus was refused. |
| 52 | //! |
| 53 | //! # References |
| 54 | //! |
| 55 | //! ITU-T H.265 (ISO/IEC 23008-2). The NAL unit header is §7.3.1.2, the sequence parameter set |
| 56 | //! §7.3.2.2, the picture parameter set §7.3.2.3, the profile-tier-level structure §7.3.3, and the |
| 57 | //! short-term reference picture sets §7.3.7. The `hvcC` record the parameter sets arrive in is |
| 58 | //! ISO/IEC 14496-15 §8.3.3. Every constant below names the clause it comes from. |
| 59 | //! |
| 60 | //! [Written with AI entirely](https://need2know.ai/entirely-ai/code)\ |
| 61 | //! Anthropic Claude |
| 62 | |
| 63 | pub mod cabac; |
| 64 | pub mod colour; |
| 65 | pub mod decode; |
| 66 | pub mod filter; |
| 67 | pub mod intra; |
| 68 | pub mod scan; |
| 69 | pub mod transform; |
| 70 | |
| 71 | pub use cabac::{ |
| 72 | Cabac, |
| 73 | Contexts, |
| 74 | Ctx, |
| 75 | Rows, |
| 76 | Set, |
| 77 | CONTEXTS, |
| 78 | }; |
| 79 | |
| 80 | use oxedyne_fe2o3_core::prelude::*; |
| 81 | |
| 82 | // The largest picture this decoder will describe, in luma samples each way. Sixteen thousand is |
| 83 | // past every camera and well inside what the level limits allow; it is a ceiling against a |
| 84 | // parameter set that is a mistake, not a limit on real photographs. |
| 85 | pub const MAX_SIDE: u32 = 16_384; |
| 86 | |
| 87 | /// NAL unit types this decoder cares about (H.265 Table 7-1). |
| 88 | pub mod nal { |
| 89 | pub const IDR_W_RADL: u8 = 19; // an IDR picture with no leading pictures, which a still is |
| 90 | pub const IDR_N_LP: u8 = 20; // the other IDR form |
| 91 | pub const VPS: u8 = 32; |
| 92 | pub const SPS: u8 = 33; |
| 93 | pub const PPS: u8 = 34; |
| 94 | } |
| 95 | |
| 96 | /// One NAL unit: what it is, and its payload with the emulation prevention undone. |
| 97 | #[derive(Clone, Debug)] |
| 98 | pub struct Unit { |
| 99 | pub kind: u8, // the type, from the two-byte NAL unit header |
| 100 | pub layer: u8, // temporal sub-layer, plus one as the header codes it |
| 101 | pub body: Vec<u8>, // after the header, every emulation prevention byte removed |
| 102 | // The payload as it arrived, escaping and all, because the entry point offsets in a slice |
| 103 | // header are counted in escaped bytes: "emulation prevention bytes that appear in the slice |
| 104 | // segment data portion of the coded slice segment NAL unit are counted as part of the slice |
| 105 | // segment data for purposes of subset identification" (§7.4.7.1). Splitting the unescaped |
| 106 | // payload at those offsets puts every row of blocks after the first escaped byte in the |
| 107 | // wrong place. |
| 108 | pub raw: Vec<u8>, |
| 109 | } |
| 110 | |
| 111 | /// What a sequence parameter set says about the pictures that follow it. |
| 112 | /// |
| 113 | /// Only the fields a still picture's decoder acts on are kept. The rest are read past, because a |
| 114 | /// parameter set is a run of variable-length codes and there is no skipping to a field without |
| 115 | /// decoding everything before it. |
| 116 | #[derive(Clone, Debug, PartialEq, Eq)] |
| 117 | pub struct Sps { |
| 118 | pub id: u8, // which set this is, as a picture parameter set names it |
| 119 | pub chroma: u8, // 0 monochrome, 1 for 4:2:0, 2 for 4:2:2, 3 for 4:4:4 |
| 120 | pub coded_w: u32, // coded width in luma samples, before the conformance window |
| 121 | pub coded_h: u32, // and coded height |
| 122 | pub width: u32, // the width the picture is meant to be shown at |
| 123 | pub height: u32, // and the height as shown |
| 124 | pub luma_bits: u8, // bits a luma sample |
| 125 | pub chroma_bits: u8, // bits a chroma sample |
| 126 | pub ctb_size: u32, // a coding tree block, in luma samples: 16, 32 or 64 |
| 127 | pub min_cb: u32, // the smallest coding block, in luma samples |
| 128 | pub min_tb: u32, // the smallest transform block, in luma samples |
| 129 | pub max_tb: u32, // and the largest |
| 130 | pub max_depth_intra: u8, // how deep the transform tree may go in an intra unit |
| 131 | pub sao: bool, // is the sample adaptive offset filter on? |
| 132 | pub pcm: bool, // may coding units carry raw samples? |
| 133 | pub strong_smoothing: bool, // the stronger intra smoothing filter, at 32 by 32 |
| 134 | // On does not mean bespoke. Every photograph in the corpus turns the scaling lists on and |
| 135 | // carries none of its own, which means the default lists apply -- and those are not flat, so |
| 136 | // a decoder that reads this as "no scaling" quantises every block wrongly and produces a |
| 137 | // picture that is recognisable and wrong. |
| 138 | pub scaling_lists: bool, // are the scaling lists in use at all? |
| 139 | pub weights: Option<Scaling>, // this sequence's own lists, or the default ones |
| 140 | // Both windows may sit off the top left corner: the conformance window usually does not and |
| 141 | // the default display window of a stabilised film always does, since it is centred in a |
| 142 | // picture coded larger than it shows. Cropping from the corner instead moves the whole |
| 143 | // picture by that offset. |
| 144 | pub show_x0: u32, // where the shown picture begins, in luma samples |
| 145 | pub show_y0: u32, // the same downwards |
| 146 | // Out of the video usability information, where a stream says how it is to be shown. A |
| 147 | // conversion into red, green and blue that guesses this wrong makes a photograph with no |
| 148 | // real black in it, or one whose blacks are crushed. |
| 149 | pub full_range: bool, // full range rather than the studio one |
| 150 | pub matrix: u8, // ISO/IEC 23091-2: 1 high definition, 5 and 6 standard |
| 151 | // A slice header carries the picture order count for every picture except an IDR, which has |
| 152 | // none to state. A film's first frame is very often a clean random access picture rather |
| 153 | // than an IDR, and a header read as though it were an IDR's is read out of step from this |
| 154 | // field on. |
| 155 | pub poc_bits: u8, // bits the count's lower part is coded in |
| 156 | // A still picture references nothing and needs none of these sets; what they are for is the |
| 157 | // slice header, which may name one or write a new one predicted from them, and either way |
| 158 | // the bits cannot be stepped over without knowing how large the set referred to is. |
| 159 | pub st_sets: Vec<(u32, u32)>, // pictures each short-term set names, negative and positive |
| 160 | pub long_term: bool, // may a slice header name long-term reference pictures? |
| 161 | pub temporal_mvp: bool, // does a slice header carry the temporal predictor flag? |
| 162 | } |
| 163 | |
| 164 | /// What a picture parameter set says about the slices that reference it. |
| 165 | #[derive(Clone, Debug, PartialEq, Eq)] |
| 166 | pub struct Pps { |
| 167 | pub id: u8, // which set this is, as a slice header names it |
| 168 | pub sps_id: u8, // which sequence parameter set it belongs to |
| 169 | pub init_qp: i32, // already offset by the 26 the syntax subtracts |
| 170 | pub cu_qp_delta: bool, // may a coding unit carry its own quantisation delta? |
| 171 | pub qp_delta_depth: u8, // how far down the quadtree a delta may be sent |
| 172 | pub cb_qp_offset: i32, // the chroma quantisation offsets |
| 173 | pub cr_qp_offset: i32, // and for the other chroma channel |
| 174 | pub transform_skip: bool, // may a block skip the transform entirely? |
| 175 | pub sign_hiding: bool, // the last coefficient's sign inferred rather than coded |
| 176 | pub transquant_bypass: bool, // an intra residual coded across the transform tree |
| 177 | pub tiles: bool, // is the picture cut into tiles? |
| 178 | pub wavefront: bool, // entropy coding synchronised at each row of blocks |
| 179 | pub deblocking: bool, // does the deblocking filter run? |
| 180 | pub slice_chroma_qp: bool, // may a slice header carry a further chroma offset? |
| 181 | // The slice header cannot be read without the count of reserved flags: they are bits to be |
| 182 | // stepped over, and stepping over the wrong number puts every field after them one place |
| 183 | // out. |
| 184 | pub extra_header_bits: u8, // reserved flags a slice header carries first |
| 185 | pub output_flag: bool, // does a slice header carry a picture output flag? |
| 186 | pub deblocking_override: bool, // may a slice header override the settings? |
| 187 | pub filter_across_slices: bool, // and therefore whether a slice carries its own flag |
| 188 | // The slice header cannot be read without this either: a segment that is not the first of |
| 189 | // its picture carries a flag saying whether it continues the header before it, and only |
| 190 | // where this says one may. |
| 191 | pub dependent_slices: bool, // may a segment continue the header before it? |
| 192 | } |
| 193 | |
| 194 | /// Splits a byte-stream of length-prefixed NAL units, as `hvcC` and `mdat` carry them. |
| 195 | /// |
| 196 | /// `length_size` comes from the configuration record and is one, two or four. A unit that runs past |
| 197 | /// the end of the buffer is a truncated file and is refused rather than decoded as far as it goes: |
| 198 | /// half a coded picture is not half a picture, it is noise. |
| 199 | pub fn split_lengthed(bytes: &[u8], length_size: usize) -> Outcome<Vec<Unit>> { |
| 200 | if !matches!(length_size, 1 | 2 | 4) { |
| 201 | return Err(err!( |
| 202 | "A NAL unit length is coded in {} bytes, and only one, two and four are legal.", |
| 203 | length_size; |
| 204 | Invalid, Input, Decode)); |
| 205 | } |
| 206 | let mut out = Vec::new(); |
| 207 | let mut at = 0usize; |
| 208 | while at + length_size <= bytes.len() { |
| 209 | let mut len = 0usize; |
| 210 | for i in 0..length_size { |
| 211 | len = (len << 8) | bytes[at + i] as usize; |
| 212 | } |
| 213 | at += length_size; |
| 214 | if len == 0 { |
| 215 | return Err(err!("A NAL unit of no length."; Invalid, Input, Decode)); |
| 216 | } |
| 217 | let end = match at.checked_add(len) { |
| 218 | Some(end) if end <= bytes.len() => end, |
| 219 | _ => return Err(err!( |
| 220 | "A NAL unit says it is {} bytes and {} remain.", len, bytes.len() - at; |
| 221 | Invalid, Input, Decode)), |
| 222 | }; |
| 223 | out.push(res!(unit(&bytes[at..end]))); |
| 224 | at = end; |
| 225 | } |
| 226 | if at != bytes.len() { |
| 227 | return Err(err!( |
| 228 | "{} bytes are left over after the last NAL unit.", bytes.len() - at; |
| 229 | Invalid, Input, Decode)); |
| 230 | } |
| 231 | Ok(out) |
| 232 | } |
| 233 | |
| 234 | /// Splits an Annex B stream, where units are separated by start codes rather than lengths. |
| 235 | /// |
| 236 | /// This is the form a parameter set arrives in inside `hvcC`, and the form a raw `.265` file takes. |
| 237 | pub fn split_annex_b(bytes: &[u8]) -> Outcome<Vec<Unit>> { |
| 238 | let mut starts: Vec<usize> = Vec::new(); |
| 239 | let mut i = 0usize; |
| 240 | while i + 3 <= bytes.len() { |
| 241 | if bytes[i] == 0 && bytes[i + 1] == 0 && bytes[i + 2] == 1 { |
| 242 | starts.push(i + 3); |
| 243 | i += 3; |
| 244 | } else { |
| 245 | i += 1; |
| 246 | } |
| 247 | } |
| 248 | let mut out = Vec::with_capacity(starts.len()); |
| 249 | for (n, from) in starts.iter().enumerate() { |
| 250 | let to = match starts.get(n + 1) { |
| 251 | // Back off the start code of the next unit, and the trailing zero a four-byte start |
| 252 | // code puts in front of it. |
| 253 | Some(next) => { |
| 254 | let mut end = next - 3; |
| 255 | if end > *from && bytes[end - 1] == 0 { |
| 256 | end -= 1; |
| 257 | } |
| 258 | end |
| 259 | }, |
| 260 | None => bytes.len(), |
| 261 | }; |
| 262 | if to > *from { |
| 263 | out.push(res!(unit(&bytes[*from..to]))); |
| 264 | } |
| 265 | } |
| 266 | Ok(out) |
| 267 | } |
| 268 | |
| 269 | /// Reads one NAL unit: its two-byte header, and its payload unescaped. |
| 270 | pub fn unit(raw: &[u8]) -> Outcome<Unit> { |
| 271 | if raw.len() < 3 { |
| 272 | return Err(err!( |
| 273 | "A NAL unit is {} bytes, and its header alone is two.", raw.len(); |
| 274 | Invalid, Input, Decode)); |
| 275 | } |
| 276 | // forbidden_zero_bit, then six bits of type, six of layer, three of temporal id (§7.3.1.2). |
| 277 | if raw[0] & 0x80 != 0 { |
| 278 | return Err(err!( |
| 279 | "A NAL unit's forbidden bit is set, so this is not an HEVC stream."; |
| 280 | Invalid, Input, Decode)); |
| 281 | } |
| 282 | Ok(Unit { |
| 283 | kind: (raw[0] >> 1) & 0x3f, |
| 284 | layer: raw[1] & 0x07, |
| 285 | body: rbsp(&raw[2..]), |
| 286 | raw: raw[2..].to_vec(), |
| 287 | }) |
| 288 | } |
| 289 | |
| 290 | /// Removes the emulation prevention bytes from a payload (§7.4.2). |
| 291 | /// |
| 292 | /// A `0x03` after two zero bytes is there only to stop the payload looking like a start code, and |
| 293 | /// is not part of the syntax. |
| 294 | pub fn rbsp(nal: &[u8]) -> Vec<u8> { |
| 295 | let mut out = Vec::with_capacity(nal.len()); |
| 296 | let mut zeros = 0usize; |
| 297 | for &b in nal { |
| 298 | if zeros >= 2 && b == 0x03 { |
| 299 | zeros = 0; |
| 300 | continue; |
| 301 | } |
| 302 | out.push(b); |
| 303 | zeros = if b == 0 { zeros + 1 } else { 0 }; |
| 304 | } |
| 305 | out |
| 306 | } |
| 307 | |
| 308 | /// Where an unescaped position sits in the payload it was unescaped from. |
| 309 | /// |
| 310 | /// Emulation prevention only ever *removes* bytes, so the escaped position is the unescaped one |
| 311 | /// plus however many were removed before it. This walks the same state machine [`rbsp`] does rather |
| 312 | /// than inverting it, because the two staying in step is the whole point. |
| 313 | pub fn escaped_at(nal: &[u8], unescaped: usize) -> usize { |
| 314 | let mut out = 0usize; |
| 315 | let mut zeros = 0usize; |
| 316 | for (i, b) in nal.iter().enumerate() { |
| 317 | if out == unescaped { |
| 318 | return i; |
| 319 | } |
| 320 | if *b == 3 && zeros >= 2 { |
| 321 | zeros = 0; |
| 322 | continue; |
| 323 | } |
| 324 | out += 1; |
| 325 | zeros = if *b == 0 { zeros + 1 } else { 0 }; |
| 326 | } |
| 327 | nal.len() |
| 328 | } |
| 329 | |
| 330 | /// The parameter sets carried in an `hvcC` decoder configuration record (ISO/IEC 14496-15 §8.3.3). |
| 331 | /// |
| 332 | /// The record's own fields describe the stream's profile and the width of the length prefixes; the |
| 333 | /// arrays at the end carry the parameter sets themselves, as Annex B payloads without start codes. |
| 334 | #[derive(Clone, Debug)] |
| 335 | pub struct Config { |
| 336 | pub length_size: usize, // bytes prefixing each NAL unit in the picture's own data |
| 337 | pub sets: Vec<Unit>, // every parameter set, in the order the record carries them |
| 338 | } |
| 339 | |
| 340 | /// Reads an `hvcC` record. |
| 341 | pub fn config(bytes: &[u8]) -> Outcome<Config> { |
| 342 | // 22 bytes of fixed fields, then a count of arrays. |
| 343 | if bytes.len() < 23 { |
| 344 | return Err(err!( |
| 345 | "A decoder configuration record is {} bytes, and its fixed fields alone are 22.", |
| 346 | bytes.len(); |
| 347 | Invalid, Input, Decode)); |
| 348 | } |
| 349 | if bytes[0] != 1 { |
| 350 | return Err(err!( |
| 351 | "A decoder configuration record of version {}, and this reads version 1.", bytes[0]; |
| 352 | Invalid, Input, Unknown)); |
| 353 | } |
| 354 | let length_size = (bytes[21] & 0x03) as usize + 1; |
| 355 | let arrays = bytes[22] as usize; |
| 356 | let mut sets = Vec::new(); |
| 357 | let mut at = 23usize; |
| 358 | for _ in 0..arrays { |
| 359 | if at + 3 > bytes.len() { |
| 360 | return Err(err!( |
| 361 | "A configuration record ends inside its array of parameter sets."; |
| 362 | Invalid, Input, Decode)); |
| 363 | } |
| 364 | let count = u16::from_be_bytes([bytes[at + 1], bytes[at + 2]]) as usize; |
| 365 | at += 3; |
| 366 | for _ in 0..count { |
| 367 | if at + 2 > bytes.len() { |
| 368 | return Err(err!( |
| 369 | "A configuration record ends inside a parameter set's length."; |
| 370 | Invalid, Input, Decode)); |
| 371 | } |
| 372 | let len = u16::from_be_bytes([bytes[at], bytes[at + 1]]) as usize; |
| 373 | at += 2; |
| 374 | let end = match at.checked_add(len) { |
| 375 | Some(end) if end <= bytes.len() => end, |
| 376 | _ => return Err(err!( |
| 377 | "A parameter set says it is {} bytes and {} remain.", |
| 378 | len, bytes.len() - at; |
| 379 | Invalid, Input, Decode)), |
| 380 | }; |
| 381 | sets.push(res!(unit(&bytes[at..end]))); |
| 382 | at = end; |
| 383 | } |
| 384 | } |
| 385 | Ok(Config { length_size, sets }) |
| 386 | } |
| 387 | |
| 388 | /// A reader of the bits of an RBSP, most significant first. |
| 389 | pub struct Bits<'a> { |
| 390 | buf: &'a [u8], |
| 391 | pos: usize, // the next bit, counted from the first bit of the first byte |
| 392 | } |
| 393 | |
| 394 | impl<'a> Bits<'a> { |
| 395 | |
| 396 | pub fn new(buf: &'a [u8]) -> Self { |
| 397 | Self { buf, pos: 0 } |
| 398 | } |
| 399 | |
| 400 | pub fn left(&self) -> usize { |
| 401 | (self.buf.len() * 8).saturating_sub(self.pos) |
| 402 | } |
| 403 | |
| 404 | /// The next `n` bits as an unsigned integer, most significant first. |
| 405 | pub fn u(&mut self, n: usize) -> Outcome<u32> { |
| 406 | if n > 32 { |
| 407 | return Err(err!("A field of {} bits was asked for, and 32 is the widest.", n; Bug)); |
| 408 | } |
| 409 | let mut v = 0u32; |
| 410 | for _ in 0..n { |
| 411 | let byte = self.pos >> 3; |
| 412 | if byte >= self.buf.len() { |
| 413 | return Err(err!( |
| 414 | "The parameter set ends after {} bits, inside a field.", self.buf.len() * 8; |
| 415 | Invalid, Input, Decode)); |
| 416 | } |
| 417 | let bit = (self.buf[byte] >> (7 - (self.pos & 7))) & 1; |
| 418 | v = (v << 1) | bit as u32; |
| 419 | self.pos += 1; |
| 420 | } |
| 421 | Ok(v) |
| 422 | } |
| 423 | |
| 424 | pub fn flag(&mut self) -> Outcome<bool> { |
| 425 | Ok(res!(self.u(1)) == 1) |
| 426 | } |
| 427 | |
| 428 | pub fn consumed(&self) -> Outcome<usize> { |
| 429 | Ok(self.pos) |
| 430 | } |
| 431 | |
| 432 | pub fn skip(&mut self, n: usize) -> Outcome<()> { |
| 433 | for _ in 0..n / 32 { |
| 434 | let _ = res!(self.u(32)); |
| 435 | } |
| 436 | let _ = res!(self.u(n % 32)); |
| 437 | Ok(()) |
| 438 | } |
| 439 | |
| 440 | /// An unsigned Exp-Golomb code, §9.2. |
| 441 | pub fn ue(&mut self) -> Outcome<u32> { |
| 442 | let mut zeros = 0usize; |
| 443 | while res!(self.u(1)) == 0 { |
| 444 | zeros += 1; |
| 445 | if zeros > 31 { |
| 446 | return Err(err!( |
| 447 | "An Exp-Golomb code is prefixed by more than 31 zeroes, which no legal value \ |
| 448 | is."; |
| 449 | Invalid, Input, Decode)); |
| 450 | } |
| 451 | } |
| 452 | if zeros == 0 { |
| 453 | return Ok(0); |
| 454 | } |
| 455 | let rest = res!(self.u(zeros)) as u64; |
| 456 | let v = (1u64 << zeros) - 1 + rest; |
| 457 | if v > u32::MAX as u64 { |
| 458 | return Err(err!( |
| 459 | "An Exp-Golomb code decodes to {}, beyond what any field holds.", v; |
| 460 | Invalid, Input, Decode)); |
| 461 | } |
| 462 | Ok(v as u32) |
| 463 | } |
| 464 | |
| 465 | /// A signed Exp-Golomb code, §9.2.2. |
| 466 | pub fn se(&mut self) -> Outcome<i32> { |
| 467 | let k = res!(self.ue()); |
| 468 | let m = ((k as i64 + 1) / 2) as i32; |
| 469 | Ok(if k % 2 == 1 { m } else { -m }) |
| 470 | } |
| 471 | } |
| 472 | |
| 473 | /// Steps over a profile, tier and level structure (§7.3.3). |
| 474 | /// |
| 475 | /// Nothing in it changes how a picture is decoded -- it says what a decoder must be capable of, and |
| 476 | /// a decoder that is about to try is going to find out. It has to be walked rather than skipped by |
| 477 | /// a byte count only in the sub-layer case, where the number of flags depends on the flags. |
| 478 | fn profile_tier_level(b: &mut Bits, profile_present: bool, max_sub_layers: usize) -> Outcome<()> { |
| 479 | if profile_present { |
| 480 | // 2 + 1 + 5 bits, 32 of compatibility flags, 48 of constraint flags. |
| 481 | res!(b.skip(8 + 32 + 48)); |
| 482 | } |
| 483 | res!(b.skip(8)); |
| 484 | if max_sub_layers == 0 { |
| 485 | return Ok(()); |
| 486 | } |
| 487 | let mut profile = [false; 8]; |
| 488 | let mut level = [false; 8]; |
| 489 | for i in 0..max_sub_layers.saturating_sub(1).min(8) { |
| 490 | profile[i] = res!(b.flag()); |
| 491 | level[i] = res!(b.flag()); |
| 492 | } |
| 493 | if max_sub_layers > 1 { |
| 494 | // The flags are padded out to eight pairs. |
| 495 | for _ in max_sub_layers.saturating_sub(1)..8 { |
| 496 | res!(b.skip(2)); |
| 497 | } |
| 498 | } |
| 499 | for i in 0..max_sub_layers.saturating_sub(1).min(8) { |
| 500 | if profile[i] { |
| 501 | res!(b.skip(8 + 32 + 48)); |
| 502 | } |
| 503 | if level[i] { |
| 504 | res!(b.skip(8)); |
| 505 | } |
| 506 | } |
| 507 | Ok(()) |
| 508 | } |
| 509 | |
| 510 | /// The weights a picture quantises each block against (§7.3.4, §7.4.5). |
| 511 | /// |
| 512 | /// Six lists a size -- one each for the three colour components, predicted from within the picture |
| 513 | /// and from another, though a still photograph only ever uses the first three. The numbers climb |
| 514 | /// away from the corner because the eye notices an error in a block's coarse detail more than in |
| 515 | /// its fine, so the fine detail is quantised harder. |
| 516 | /// |
| 517 | /// A sequence that turns the lists on and carries none of its own takes the default ones, which are |
| 518 | /// not flat; a decoder that reads "on" as "no scaling" quantises every block wrongly and produces a |
| 519 | /// picture that is recognisable and wrong. |
| 520 | #[derive(Clone, Debug, PartialEq, Eq)] |
| 521 | pub struct Scaling { |
| 522 | pub list: [[[u8; 64]; 6]; 4], // ScalingList[sizeId][matrixId][i], in diagonal scan order |
| 523 | pub dc: [[u8; 6]; 2], // the corner at the two largest sizes, coded on its own |
| 524 | } |
| 525 | |
| 526 | impl Scaling { |
| 527 | |
| 528 | /// The default lists, which is what a sequence carrying none of its own means. |
| 529 | pub fn default_lists() -> Self { |
| 530 | let mut out = Self { list: [[[16u8; 64]; 6]; 4], dc: [[16u8; 6]; 2] }; |
| 531 | for size in 1..4 { |
| 532 | for id in 0..6 { |
| 533 | let from = crate::hevc::transform::DEFAULT_LIST[(id >= 3) as usize]; |
| 534 | out.list[size][id] = from; |
| 535 | } |
| 536 | } |
| 537 | out |
| 538 | } |
| 539 | |
| 540 | /// One weight, by size, matrix and position within the block (equations 7-44 to 7-49). |
| 541 | /// |
| 542 | /// The sixteen and thirty-two sample matrices are the eight-sample one with each of its values |
| 543 | /// covering two or four samples each way, and a corner of their own. |
| 544 | pub fn factor(&self, log2: u32, matrix: usize, x: usize, y: usize, raster: &[u8; 64]) -> i32 { |
| 545 | match log2 { |
| 546 | 2 => raster[(y & 3) * 4 + (x & 3)] as i32, |
| 547 | 3 => raster[y * 8 + x] as i32, |
| 548 | _ => { |
| 549 | if x == 0 && y == 0 { |
| 550 | return self.dc[(log2 - 4) as usize][matrix] as i32; |
| 551 | } |
| 552 | let shrink = log2 - 3; |
| 553 | raster[(y >> shrink) * 8 + (x >> shrink)] as i32 |
| 554 | }, |
| 555 | } |
| 556 | } |
| 557 | |
| 558 | /// One list laid out in raster order rather than in the diagonal scan's. |
| 559 | pub fn raster(&self, log2: u32, matrix: usize) -> [u8; 64] { |
| 560 | let size_id = (log2 - 2).min(3) as usize; |
| 561 | let side = if size_id == 0 { 4 } else { 8 }; |
| 562 | let list = &self.list[size_id][matrix]; |
| 563 | let mut out = [16u8; 64]; |
| 564 | for (i, (x, y)) in crate::hevc::scan::positions(side, crate::hevc::scan::Order::Diagonal) |
| 565 | .iter() |
| 566 | .enumerate() |
| 567 | { |
| 568 | out[*y as usize * side + *x as usize] = list[i]; |
| 569 | } |
| 570 | out |
| 571 | } |
| 572 | } |
| 573 | |
| 574 | /// Reads a scaling list (§7.3.4). |
| 575 | /// |
| 576 | /// A list is either coded outright as a chain of differences, or taken from an earlier list in the |
| 577 | /// same set, or -- where it names itself as its own source -- from the default. |
| 578 | fn scaling_list(b: &mut Bits) -> Outcome<Scaling> { |
| 579 | let mut out = Scaling::default_lists(); |
| 580 | let defaults = Scaling::default_lists(); |
| 581 | for size in 0..4usize { |
| 582 | let mut id = 0usize; |
| 583 | while id < 6 { |
| 584 | let coefficients = 64usize.min(1 << (4 + (size << 1))); |
| 585 | if !res!(b.flag()) { |
| 586 | // Taken from another list rather than coded. A delta of nought means the default, |
| 587 | // which is the one case where "predicted from" does not mean "copied from". |
| 588 | let delta = res!(b.ue()) as usize; |
| 589 | if delta == 0 { |
| 590 | out.list[size][id] = defaults.list[size][id]; |
| 591 | if size > 1 { |
| 592 | out.dc[size - 2][id] = 16; |
| 593 | } |
| 594 | } else { |
| 595 | let from = id.saturating_sub(delta * if size == 3 { 3 } else { 1 }); |
| 596 | out.list[size][id] = out.list[size][from]; |
| 597 | if size > 1 { |
| 598 | out.dc[size - 2][id] = out.dc[size - 2][from]; |
| 599 | } |
| 600 | } |
| 601 | } else { |
| 602 | let mut next = 8i32; |
| 603 | if size > 1 { |
| 604 | let dc = res!(b.se()) + 8; |
| 605 | if !(1..=255).contains(&dc) { |
| 606 | return Err(err!( |
| 607 | "A scaling list's corner value is {}, outside 1 to 255.", dc; |
| 608 | Invalid, Input, Decode)); |
| 609 | } |
| 610 | out.dc[size - 2][id] = dc as u8; |
| 611 | next = dc; |
| 612 | } |
| 613 | for i in 0..coefficients { |
| 614 | let delta = res!(b.se()); |
| 615 | next = (next + delta + 256).rem_euclid(256); |
| 616 | out.list[size][id][i] = next as u8; |
| 617 | } |
| 618 | } |
| 619 | // The 32 by 32 lists come in twos rather than sixes. |
| 620 | id += if size == 3 { 3 } else { 1 }; |
| 621 | } |
| 622 | } |
| 623 | // A 32 by 32 chroma list does not exist below 4:4:4, but the arrays are square; filling the |
| 624 | // gaps from luma keeps a lookup by matrix identifier from finding sixteens. |
| 625 | for id in [1usize, 2, 4, 5] { |
| 626 | let from = if id < 3 { 0 } else { 3 }; |
| 627 | out.list[3][id] = out.list[3][from]; |
| 628 | out.dc[1][id] = out.dc[1][from]; |
| 629 | } |
| 630 | Ok(out) |
| 631 | } |
| 632 | |
| 633 | /// Steps over one short-term reference picture set (§7.3.7). |
| 634 | /// |
| 635 | /// A still picture references nothing, so no *picture* is kept -- but how many the set names is, |
| 636 | /// because the next set may be coded as a difference from this one and a slice header may be coded |
| 637 | /// as a difference from any of them, and neither can be stepped over without the count. |
| 638 | fn short_term_ref_pic_set(b: &mut Bits, idx: usize, count: usize, previous: &mut Vec<(u32, u32)>) |
| 639 | -> Outcome<()> |
| 640 | { |
| 641 | let mut predicted = false; |
| 642 | if idx != 0 { |
| 643 | predicted = res!(b.flag()); |
| 644 | } |
| 645 | if predicted { |
| 646 | // Which earlier set this one is a difference from. Only a set written in a slice header |
| 647 | // says so; a set in the sequence parameter set is always a difference from the one before |
| 648 | // it (§7.4.8). |
| 649 | let mut back = 1usize; |
| 650 | if idx == count { |
| 651 | back = res!(b.ue()) as usize + 1; |
| 652 | } |
| 653 | let _delta_rps_sign = res!(b.flag()); |
| 654 | let _abs_delta_rps = res!(b.ue()); |
| 655 | let (negative, positive) = match idx.checked_sub(back).and_then(|at| previous.get(at)) { |
| 656 | Some(pair) => *pair, |
| 657 | None => return Err(err!( |
| 658 | "A reference picture set is coded as a difference from set {} of {}, which is not \ |
| 659 | there.", idx as i64 - back as i64, previous.len(); |
| 660 | Invalid, Input, Decode)), |
| 661 | }; |
| 662 | // One flag pair for each picture of the set referred to, and one for the picture that set |
| 663 | // is itself relative to. The ones kept are what this set names, which is what the next |
| 664 | // difference will be measured against. |
| 665 | let mut kept = 0u32; |
| 666 | for _ in 0..(negative + positive + 1) { |
| 667 | let used = res!(b.flag()); |
| 668 | let mut keep = used; |
| 669 | if !used { |
| 670 | keep = res!(b.flag()); |
| 671 | } |
| 672 | if keep { |
| 673 | kept += 1; |
| 674 | } |
| 675 | } |
| 676 | previous.push((kept, 0)); |
| 677 | return Ok(()); |
| 678 | } |
| 679 | let negative = res!(b.ue()); |
| 680 | let positive = res!(b.ue()); |
| 681 | if negative > 64 || positive > 64 { |
| 682 | return Err(err!( |
| 683 | "A reference picture set names {} and {} pictures, and 64 is the most either may be.", |
| 684 | negative, positive; |
| 685 | Invalid, Input, Decode)); |
| 686 | } |
| 687 | for _ in 0..negative { |
| 688 | let _delta = res!(b.ue()); |
| 689 | let _used = res!(b.flag()); |
| 690 | } |
| 691 | for _ in 0..positive { |
| 692 | let _delta = res!(b.ue()); |
| 693 | let _used = res!(b.flag()); |
| 694 | } |
| 695 | previous.push((negative, positive)); |
| 696 | Ok(()) |
| 697 | } |
| 698 | |
| 699 | /// Reads a sequence parameter set (§7.3.2.2). |
| 700 | pub fn sps(body: &[u8]) -> Outcome<Sps> { |
| 701 | let mut b = Bits::new(body); |
| 702 | let _vps_id = res!(b.u(4)); |
| 703 | let max_sub_layers = res!(b.u(3)) as usize + 1; |
| 704 | let _temporal_id_nesting = res!(b.flag()); |
| 705 | res!(profile_tier_level(&mut b, true, max_sub_layers)); |
| 706 | let id = res!(b.ue()); |
| 707 | if id > 15 { |
| 708 | return Err(err!( |
| 709 | "A sequence parameter set numbered {}, and 15 is the highest.", id; |
| 710 | Invalid, Input, Decode)); |
| 711 | } |
| 712 | let chroma = res!(b.ue()); |
| 713 | if chroma > 3 { |
| 714 | return Err(err!( |
| 715 | "A chroma format of {}, and 3 is the highest.", chroma; Invalid, Input, Decode)); |
| 716 | } |
| 717 | if chroma == 3 { |
| 718 | let _separate_colour_plane = res!(b.flag()); |
| 719 | } |
| 720 | let coded_w = res!(b.ue()); |
| 721 | let coded_h = res!(b.ue()); |
| 722 | if coded_w == 0 || coded_h == 0 || coded_w > MAX_SIDE || coded_h > MAX_SIDE { |
| 723 | return Err(err!( |
| 724 | "A sequence parameter set codes a picture of {} by {}.", coded_w, coded_h; |
| 725 | Invalid, Input, Decode)); |
| 726 | } |
| 727 | // The conformance window trims the coded picture down to what is shown, in units of the chroma |
| 728 | // sampling: a 1920 by 1080 picture is coded as 1920 by 1088 and trimmed by four rows. |
| 729 | let (mut left, mut right, mut top, mut bottom) = (0u32, 0u32, 0u32, 0u32); |
| 730 | if res!(b.flag()) { |
| 731 | left = res!(b.ue()); |
| 732 | right = res!(b.ue()); |
| 733 | top = res!(b.ue()); |
| 734 | bottom = res!(b.ue()); |
| 735 | } |
| 736 | let (sub_w, sub_h) = match chroma { |
| 737 | 1 => (2u32, 2u32), |
| 738 | 2 => (2, 1), |
| 739 | _ => (1, 1), |
| 740 | }; |
| 741 | let trim_x = left.saturating_add(right).saturating_mul(sub_w); |
| 742 | let trim_y = top.saturating_add(bottom).saturating_mul(sub_h); |
| 743 | if trim_x >= coded_w || trim_y >= coded_h { |
| 744 | return Err(err!( |
| 745 | "A conformance window trims {} by {} from a picture of {} by {}.", |
| 746 | trim_x, trim_y, coded_w, coded_h; |
| 747 | Invalid, Input, Decode)); |
| 748 | } |
| 749 | let luma_bits = res!(b.ue()) as u8 + 8; |
| 750 | let chroma_bits = res!(b.ue()) as u8 + 8; |
| 751 | if luma_bits > 16 || chroma_bits > 16 { |
| 752 | return Err(err!( |
| 753 | "A sample of {} bits, and 16 is the most this decoder reads.", luma_bits.max(chroma_bits); |
| 754 | Invalid, Input, Unknown)); |
| 755 | } |
| 756 | let poc_bits = res!(b.ue()) as u8 + 4; |
| 757 | if poc_bits > 16 { |
| 758 | return Err(err!( |
| 759 | "A picture order count of {} bits, and 16 is the most.", poc_bits; |
| 760 | Invalid, Input, Decode)); |
| 761 | } |
| 762 | // The ordering information is given either once for the highest sub-layer or once for each. |
| 763 | let for_each = res!(b.flag()); |
| 764 | let first = if for_each { 0 } else { max_sub_layers - 1 }; |
| 765 | for _ in first..max_sub_layers { |
| 766 | let _max_dec_pic_buffering = res!(b.ue()); |
| 767 | let _num_reorder = res!(b.ue()); |
| 768 | let _max_latency = res!(b.ue()); |
| 769 | } |
| 770 | let min_cb = 1u32 << (res!(b.ue()) + 3); |
| 771 | let ctb_size = min_cb << res!(b.ue()); |
| 772 | let min_tb = 1u32 << (res!(b.ue()) + 2); |
| 773 | let max_tb = min_tb << res!(b.ue()); |
| 774 | if !matches!(ctb_size, 16 | 32 | 64) || min_cb < 8 || max_tb > 32 || min_tb < 4 { |
| 775 | return Err(err!( |
| 776 | "A block geometry of ctb {}, min cb {}, tb {} to {}, which no legal stream has.", |
| 777 | ctb_size, min_cb, min_tb, max_tb; |
| 778 | Invalid, Input, Decode)); |
| 779 | } |
| 780 | let _max_depth_inter = res!(b.ue()); |
| 781 | let max_depth_intra = res!(b.ue()) as u8; |
| 782 | let scaling_lists = res!(b.flag()); |
| 783 | let mut weights = None; |
| 784 | if scaling_lists { |
| 785 | weights = Some(if res!(b.flag()) { |
| 786 | res!(scaling_list(&mut b)) |
| 787 | } else { |
| 788 | Scaling::default_lists() |
| 789 | }); |
| 790 | } |
| 791 | let _amp = res!(b.flag()); |
| 792 | let sao = res!(b.flag()); |
| 793 | let pcm = res!(b.flag()); |
| 794 | if pcm { |
| 795 | let _pcm_luma_bits = res!(b.u(4)); |
| 796 | let _pcm_chroma_bits = res!(b.u(4)); |
| 797 | let _log2_min_pcm_cb = res!(b.ue()); |
| 798 | let _log2_diff_pcm_cb = res!(b.ue()); |
| 799 | let _pcm_loop_filter_disabled = res!(b.flag()); |
| 800 | } |
| 801 | let short_term_sets = res!(b.ue()) as usize; |
| 802 | if short_term_sets > 64 { |
| 803 | return Err(err!( |
| 804 | "A sequence parameter set carries {} reference picture sets, and 64 is the most.", |
| 805 | short_term_sets; |
| 806 | Invalid, Input, Decode)); |
| 807 | } |
| 808 | let mut previous: Vec<(u32, u32)> = Vec::with_capacity(short_term_sets); |
| 809 | for i in 0..short_term_sets { |
| 810 | res!(short_term_ref_pic_set(&mut b, i, short_term_sets, &mut previous)); |
| 811 | } |
| 812 | let long_term_present = res!(b.flag()); |
| 813 | if long_term_present { |
| 814 | let long_term = res!(b.ue()) as usize; |
| 815 | if long_term > 32 { |
| 816 | return Err(err!( |
| 817 | "A sequence parameter set carries {} long-term reference pictures.", long_term; |
| 818 | Invalid, Input, Decode)); |
| 819 | } |
| 820 | for _ in 0..long_term { |
| 821 | let bits = (res!(b.ue()) % 32) as usize; |
| 822 | let _ = bits; |
| 823 | // The poc is coded in log2_max_poc bits, which was read past above; a still picture |
| 824 | // has none of these, and a stream that does is not one this decoder will be handed. |
| 825 | return Err(err!( |
| 826 | "A sequence parameter set carries long-term reference pictures, which a still \ |
| 827 | picture does not have."; |
| 828 | Invalid, Input, Unknown)); |
| 829 | } |
| 830 | } |
| 831 | let temporal_mvp = res!(b.flag()); |
| 832 | let strong_smoothing = res!(b.flag()); |
| 833 | // The video usability information, which is where a stream says how it is to be *shown*: which |
| 834 | // weights its colour was coded against, whether its samples run the full range, and -- the one |
| 835 | // that changes the picture's size -- the default display window. |
| 836 | // |
| 837 | // **A phone's stabilised film carries one.** Stabilisation works by coding a picture larger |
| 838 | // than it shows and moving the window about inside it, and the window is written here. A |
| 839 | // decoder that ignores it hands back the wobbly margin as though it were part of the film, |
| 840 | // about nine per cent wider and taller than every player shows. |
| 841 | let (mut full_range, mut matrix) = (false, 2u8); |
| 842 | let (mut show_x, mut show_y) = (0u32, 0u32); |
| 843 | let (mut show_x0, mut show_y0) = (0u32, 0u32); |
| 844 | if res!(b.flag()) { |
| 845 | if res!(b.flag()) { |
| 846 | // The sample aspect ratio, read past: a picture is drawn at the size it is coded and |
| 847 | // stretching it is the caller's business. |
| 848 | let idc = res!(b.u(8)); |
| 849 | if idc == 255 { |
| 850 | let _sar_w = res!(b.u(16)); |
| 851 | let _sar_h = res!(b.u(16)); |
| 852 | } |
| 853 | } |
| 854 | if res!(b.flag()) { |
| 855 | let _overscan_appropriate = res!(b.flag()); |
| 856 | } |
| 857 | if res!(b.flag()) { |
| 858 | let _video_format = res!(b.u(3)); |
| 859 | full_range = res!(b.flag()); |
| 860 | if res!(b.flag()) { |
| 861 | let _primaries = res!(b.u(8)); |
| 862 | let _transfer = res!(b.u(8)); |
| 863 | matrix = res!(b.u(8)) as u8; |
| 864 | } |
| 865 | } |
| 866 | if res!(b.flag()) { |
| 867 | let _chroma_loc_top = res!(b.ue()); |
| 868 | let _chroma_loc_bottom = res!(b.ue()); |
| 869 | } |
| 870 | let _neutral_chroma = res!(b.flag()); |
| 871 | let _field_seq = res!(b.flag()); |
| 872 | let _frame_field_info = res!(b.flag()); |
| 873 | if res!(b.flag()) { |
| 874 | let dw_left = res!(b.ue()); |
| 875 | let dw_right = res!(b.ue()); |
| 876 | let dw_top = res!(b.ue()); |
| 877 | let dw_bottom = res!(b.ue()); |
| 878 | show_x = dw_left.saturating_add(dw_right).saturating_mul(sub_w); |
| 879 | show_y = dw_top.saturating_add(dw_bottom).saturating_mul(sub_h); |
| 880 | show_x0 = dw_left.saturating_mul(sub_w); |
| 881 | show_y0 = dw_top.saturating_mul(sub_h); |
| 882 | } |
| 883 | // Nothing after the window is read: the timing information, the bitstream restrictions and |
| 884 | // the hypothetical reference decoder say nothing about the samples. |
| 885 | } |
| 886 | let width = coded_w - trim_x; |
| 887 | let height = coded_h - trim_y; |
| 888 | if show_x >= width || show_y >= height { |
| 889 | return Err(err!( |
| 890 | "A default display window trims {} by {} from a picture of {} by {}.", |
| 891 | show_x, show_y, width, height; |
| 892 | Invalid, Input, Range)); |
| 893 | } |
| 894 | Ok(Sps { |
| 895 | id: id as u8, |
| 896 | chroma: chroma as u8, |
| 897 | coded_w, |
| 898 | coded_h, |
| 899 | width: width - show_x, |
| 900 | height: height - show_y, |
| 901 | show_x0: left.saturating_mul(sub_w) + show_x0, |
| 902 | show_y0: top.saturating_mul(sub_h) + show_y0, |
| 903 | luma_bits, |
| 904 | chroma_bits, |
| 905 | ctb_size, |
| 906 | min_cb, |
| 907 | min_tb, |
| 908 | max_tb, |
| 909 | max_depth_intra, |
| 910 | sao, |
| 911 | pcm, |
| 912 | strong_smoothing, |
| 913 | scaling_lists, |
| 914 | weights, |
| 915 | poc_bits, |
| 916 | full_range, |
| 917 | matrix, |
| 918 | st_sets: previous, |
| 919 | long_term: long_term_present, |
| 920 | temporal_mvp, |
| 921 | }) |
| 922 | } |
| 923 | |
| 924 | /// Reads a picture parameter set (§7.3.2.3). |
| 925 | pub fn pps(body: &[u8]) -> Outcome<Pps> { |
| 926 | let mut b = Bits::new(body); |
| 927 | let id = res!(b.ue()); |
| 928 | let sps_id = res!(b.ue()); |
| 929 | if id > 63 || sps_id > 15 { |
| 930 | return Err(err!( |
| 931 | "A picture parameter set numbered {} against sequence set {}.", id, sps_id; |
| 932 | Invalid, Input, Decode)); |
| 933 | } |
| 934 | let dependent_slices = res!(b.flag()); |
| 935 | let output_flag = res!(b.flag()); |
| 936 | let extra_header_bits = res!(b.u(3)) as u8; |
| 937 | let sign_hiding = res!(b.flag()); |
| 938 | let _cabac_init_present = res!(b.flag()); |
| 939 | let _num_ref_idx_l0 = res!(b.ue()); |
| 940 | let _num_ref_idx_l1 = res!(b.ue()); |
| 941 | let init_qp = res!(b.se()) + 26; |
| 942 | let _constrained_intra_pred = res!(b.flag()); |
| 943 | let transform_skip = res!(b.flag()); |
| 944 | let cu_qp_delta = res!(b.flag()); |
| 945 | let qp_delta_depth = if cu_qp_delta { res!(b.ue()) as u8 } else { 0 }; |
| 946 | let cb_qp_offset = res!(b.se()); |
| 947 | let cr_qp_offset = res!(b.se()); |
| 948 | let slice_chroma_qp = res!(b.flag()); |
| 949 | let _weighted_pred = res!(b.flag()); |
| 950 | let _weighted_bipred = res!(b.flag()); |
| 951 | let transquant_bypass = res!(b.flag()); |
| 952 | let tiles = res!(b.flag()); |
| 953 | let wavefront = res!(b.flag()); |
| 954 | if tiles { |
| 955 | // The geometry of the tiles is read past rather than kept: what this decoder needs from a |
| 956 | // tiled picture is to know it is one, and to say so. |
| 957 | let columns = res!(b.ue()) as usize; |
| 958 | let rows = res!(b.ue()) as usize; |
| 959 | if columns > 1024 || rows > 1024 { |
| 960 | return Err(err!( |
| 961 | "A picture in {} by {} tiles.", columns + 1, rows + 1; Invalid, Input, Decode)); |
| 962 | } |
| 963 | if !res!(b.flag()) { |
| 964 | for _ in 0..columns { |
| 965 | let _width = res!(b.ue()); |
| 966 | } |
| 967 | for _ in 0..rows { |
| 968 | let _height = res!(b.ue()); |
| 969 | } |
| 970 | } |
| 971 | let _loop_filter_across_tiles = res!(b.flag()); |
| 972 | } |
| 973 | let filter_across_slices = res!(b.flag()); |
| 974 | let mut deblocking = true; |
| 975 | let mut deblocking_override = false; |
| 976 | if res!(b.flag()) { |
| 977 | deblocking_override = res!(b.flag()); |
| 978 | deblocking = !res!(b.flag()); |
| 979 | if deblocking { |
| 980 | let _beta_offset = res!(b.se()); |
| 981 | let _tc_offset = res!(b.se()); |
| 982 | } |
| 983 | } |
| 984 | Ok(Pps { |
| 985 | id: id as u8, |
| 986 | sps_id: sps_id as u8, |
| 987 | init_qp, |
| 988 | cu_qp_delta, |
| 989 | qp_delta_depth, |
| 990 | cb_qp_offset, |
| 991 | cr_qp_offset, |
| 992 | transform_skip, |
| 993 | sign_hiding, |
| 994 | transquant_bypass, |
| 995 | tiles, |
| 996 | wavefront, |
| 997 | deblocking, |
| 998 | slice_chroma_qp, |
| 999 | extra_header_bits, |
| 1000 | output_flag, |
| 1001 | deblocking_override, |
| 1002 | filter_across_slices, |
| 1003 | dependent_slices, |
| 1004 | }) |
| 1005 | } |
| 1006 | |
| 1007 | /// What a slice segment header says, for the one kind of slice a still picture has. |
| 1008 | /// |
| 1009 | /// A picture is one slice and the slice is intra, so most of the syntax -- reference lists, |
| 1010 | /// weighted prediction, temporal motion vectors -- is not reached at all. What matters here is |
| 1011 | /// where the header **ends**: the arithmetic decoder starts at the next byte boundary after it, and |
| 1012 | /// a header read one bit short starts the whole of the rest of the decode in the wrong place. |
| 1013 | #[derive(Clone, Debug, PartialEq, Eq)] |
| 1014 | pub struct Slice { |
| 1015 | pub first: bool, // is this segment the first of its picture? |
| 1016 | // Nought for the first segment of a picture, which is every segment of a picture that is |
| 1017 | // one slice -- which every photograph is and many films are not. |
| 1018 | pub address: u32, // the coding tree block it begins at, raster order from nought |
| 1019 | pub across_slices: bool, // the picture parameter set's answer where absent (§7.4.7.1) |
| 1020 | pub pps_id: u8, // which picture parameter set the slice references |
| 1021 | pub kind: u8, // 2 is intra, and this decoder reads no other |
| 1022 | pub qp: i32, // the quantisation parameter this slice starts at |
| 1023 | pub sao_luma: bool, // does the sample adaptive offset run on luma here? |
| 1024 | pub sao_chroma: bool, // and on chroma? |
| 1025 | pub data_at: usize, // the header's end rounded up to a byte, where §9.3.1 starts |
| 1026 | pub deblocking: bool, // does the deblocking filter run on this slice? |
| 1027 | pub cb_qp_offset: i32, // what this slice adds to the picture's chroma offsets |
| 1028 | pub cr_qp_offset: i32, // and for the other chroma component |
| 1029 | // One a row of coding tree blocks, under wavefront coding, which is what every photograph |
| 1030 | // measured uses. |
| 1031 | pub entries: Vec<u64>, // where each piece begins, as the length of the one before |
| 1032 | } |
| 1033 | |
| 1034 | /// Which picture parameter set a slice names, read without the set itself. |
| 1035 | /// |
| 1036 | /// The identifier is the third element of the header and none of the three before it depends on a |
| 1037 | /// parameter set, so it can be had before choosing one -- which is the point: a caller holding |
| 1038 | /// several sets has to know which it is being asked for. It sits before the segment address, so |
| 1039 | /// this reads the same three elements whether or not the segment is the first of its picture. |
| 1040 | pub fn slice_pps_id(body: &[u8]) -> Outcome<u8> { |
| 1041 | let mut b = Bits::new(body); |
| 1042 | let _first = res!(b.flag()); |
| 1043 | // Only an IRAP picture carries this flag, and every still is one. |
| 1044 | let _no_output_of_prior_pics = res!(b.flag()); |
| 1045 | Ok(res!(b.ue()) as u8) |
| 1046 | } |
| 1047 | |
| 1048 | /// Reads a slice segment header (§7.3.6.1). |
| 1049 | /// |
| 1050 | /// Only the independent, intra case: a **dependent** slice segment carries no header of its own but |
| 1051 | /// continues the one before it, and is refused by name. |
| 1052 | /// |
| 1053 | /// A segment that is not the first of its picture carries the coding tree block it begins at, in as |
| 1054 | /// many bits as it takes to count the picture's blocks -- which is why the sequence parameter set is |
| 1055 | /// needed to read a header at all. |
| 1056 | pub fn slice(body: &[u8], sps: &Sps, pps: &Pps) -> Outcome<Slice> { |
| 1057 | slice_of(nal::IDR_W_RADL, body, sps, pps) |
| 1058 | } |
| 1059 | |
| 1060 | /// The same, for a slice of a picture that may not be an IDR. |
| 1061 | /// |
| 1062 | /// **A film's first frame very often is not one.** A clean random access picture opens a stream |
| 1063 | /// just as an IDR does and is decoded exactly as one -- it references nothing before itself -- but |
| 1064 | /// its slice header carries the picture order count and the reference picture set that an IDR's |
| 1065 | /// does not, because the pictures *after* it may reference what it names. A header read as though |
| 1066 | /// it were an IDR's is read out of step from that field onwards, and what comes out is a plausible |
| 1067 | /// number of entry points and a picture of noise. |
| 1068 | /// |
| 1069 | /// `kind` is the NAL unit type, which is the only thing that says which of the two this is. |
| 1070 | pub fn slice_of(kind: u8, body: &[u8], sps: &Sps, pps: &Pps) -> Outcome<Slice> { |
| 1071 | let mut b = Bits::new(body); |
| 1072 | let first = res!(b.flag()); |
| 1073 | // Only an IRAP picture carries this flag, and every still is one. |
| 1074 | let _no_output_of_prior_pics = res!(b.flag()); |
| 1075 | let pps_id = res!(b.ue()); |
| 1076 | let mut address = 0u32; |
| 1077 | if !first { |
| 1078 | if pps.dependent_slices && res!(b.flag()) { |
| 1079 | return Err(err!( |
| 1080 | "A dependent slice segment, which carries no header of its own but continues the \ |
| 1081 | one before it."; Unimplemented)); |
| 1082 | } |
| 1083 | // As many bits as it takes to count the picture's coding tree blocks (§7.4.7.1). |
| 1084 | let ctb = sps.ctb_size.max(1); |
| 1085 | let blocks = ((sps.coded_w + ctb - 1) / ctb) as u64 * ((sps.coded_h + ctb - 1) / ctb) as u64; |
| 1086 | let mut width = 0usize; |
| 1087 | while (1u64 << width) < blocks { |
| 1088 | width += 1; |
| 1089 | } |
| 1090 | address = res!(b.u(width)) as u32; |
| 1091 | if address as u64 >= blocks { |
| 1092 | return Err(err!( |
| 1093 | "A slice segment begins at block {} of a picture holding {}.", address, blocks; |
| 1094 | Invalid, Input, Decode)); |
| 1095 | } |
| 1096 | } |
| 1097 | if pps_id as u8 != pps.id { |
| 1098 | return Err(err!( |
| 1099 | "A slice references picture parameter set {} and the one in hand is {}.", |
| 1100 | pps_id, pps.id; |
| 1101 | Invalid, Input, Missing)); |
| 1102 | } |
| 1103 | // Reserved, and to be stepped over rather than understood. Stepping over the wrong number of |
| 1104 | // them puts every field after them one place out, which is why the count is carried here from |
| 1105 | // the picture parameter set rather than assumed to be zero. They come **before** the slice |
| 1106 | // type (§7.3.6.1). |
| 1107 | res!(b.skip(pps.extra_header_bits as usize)); |
| 1108 | let slice_kind = res!(b.ue()); |
| 1109 | if slice_kind != 2 { |
| 1110 | return Err(err!( |
| 1111 | "A slice of type {}, and a still picture's slices are all intra (type 2).", slice_kind; |
| 1112 | Invalid, Input, Unknown)); |
| 1113 | } |
| 1114 | if pps.output_flag { |
| 1115 | let _pic_output_flag = res!(b.flag()); |
| 1116 | } |
| 1117 | // What an IDR does not carry, and everything else does: where this picture sits in output |
| 1118 | // order, and which pictures the ones after it may reference. |
| 1119 | if kind != nal::IDR_W_RADL && kind != nal::IDR_N_LP { |
| 1120 | let _poc_lsb = res!(b.u(sps.poc_bits as usize)); |
| 1121 | let from_sps = res!(b.flag()); |
| 1122 | if !from_sps { |
| 1123 | // A set of its own, written here and coded as a difference from one of the sequence's. |
| 1124 | let mut sets = sps.st_sets.clone(); |
| 1125 | let count = sets.len(); |
| 1126 | res!(short_term_ref_pic_set(&mut b, count, count, &mut sets)); |
| 1127 | } else if sps.st_sets.len() > 1 { |
| 1128 | // As many bits as it takes to count them (§7.4.7.1). |
| 1129 | let mut width = 0usize; |
| 1130 | while (1usize << width) < sps.st_sets.len() { |
| 1131 | width += 1; |
| 1132 | } |
| 1133 | let _which = res!(b.u(width)); |
| 1134 | } |
| 1135 | if sps.long_term { |
| 1136 | // A sequence carrying long-term reference pictures is refused where it is read, so |
| 1137 | // reaching this means the flag is set and the sequence names none of them. |
| 1138 | let _num_long_term_pics = res!(b.ue()); |
| 1139 | return Err(err!( |
| 1140 | "A slice names long-term reference pictures, which a picture decoded on its own \ |
| 1141 | has no use for and this reader does not follow."; Unimplemented)); |
| 1142 | } |
| 1143 | if sps.temporal_mvp { |
| 1144 | let _temporal_mvp = res!(b.flag()); |
| 1145 | } |
| 1146 | } |
| 1147 | let mut sao_luma = false; |
| 1148 | let mut sao_chroma = false; |
| 1149 | if sps.sao { |
| 1150 | sao_luma = res!(b.flag()); |
| 1151 | if sps.chroma != 0 { |
| 1152 | sao_chroma = res!(b.flag()); |
| 1153 | } |
| 1154 | } |
| 1155 | let qp = pps.init_qp + res!(b.se()); |
| 1156 | if qp < -(6 * (sps.luma_bits as i32 - 8)) || qp > 51 { |
| 1157 | return Err(err!( |
| 1158 | "A slice starts at a quantisation parameter of {}, outside the legal range.", qp; |
| 1159 | Invalid, Input, Decode)); |
| 1160 | } |
| 1161 | // The chroma offsets a slice may add to the picture's own. Kept rather than stepped over: |
| 1162 | // they go into the chroma quantisation parameter of every block, so a picture whose slice |
| 1163 | // carries one and whose decoder ignores it comes out with the wrong colour saturation. |
| 1164 | let (mut cb_offset, mut cr_offset) = (0i32, 0i32); |
| 1165 | if pps.slice_chroma_qp { |
| 1166 | cb_offset = res!(b.se()); |
| 1167 | cr_offset = res!(b.se()); |
| 1168 | } |
| 1169 | let mut deblocking = pps.deblocking; |
| 1170 | if pps.deblocking_override && res!(b.flag()) { |
| 1171 | deblocking = !res!(b.flag()); |
| 1172 | if deblocking { |
| 1173 | let _beta = res!(b.se()); |
| 1174 | let _tc = res!(b.se()); |
| 1175 | } |
| 1176 | } |
| 1177 | // Whether the loop filters run across this slice's boundaries. Where the header does not carry |
| 1178 | // it, the picture parameter set's answer stands (§7.4.7.1). |
| 1179 | let mut across_slices = pps.filter_across_slices; |
| 1180 | if pps.filter_across_slices && (sao_luma || sao_chroma || deblocking) { |
| 1181 | across_slices = res!(b.flag()); |
| 1182 | } |
| 1183 | // Where the picture is cut up for parallel decoding, the header says where each piece begins. |
| 1184 | // |
| 1185 | // **A still photograph out of a phone is coded this way.** Every one of the 359 HEIC files |
| 1186 | // measured sets `entropy_coding_sync_enabled_flag`, which is wavefront coding: the arithmetic |
| 1187 | // decoder is reset at the start of every row of coding tree blocks, from the state saved after |
| 1188 | // the second block of the row above. So this is not an exotic case to be refused -- it is the |
| 1189 | // case, and the offsets below are how the rows are found. |
| 1190 | let mut entries: Vec<u64> = Vec::new(); |
| 1191 | if pps.tiles || pps.wavefront { |
| 1192 | let count = res!(b.ue()) as usize; |
| 1193 | if count > 4096 { |
| 1194 | return Err(err!( |
| 1195 | "A slice names {} entry points, and no picture this decoder reads has so many.", |
| 1196 | count; |
| 1197 | Invalid, Input, Decode)); |
| 1198 | } |
| 1199 | if count > 0 { |
| 1200 | let width = res!(b.ue()) as usize + 1; |
| 1201 | if width > 32 { |
| 1202 | return Err(err!( |
| 1203 | "An entry point offset of {} bits, and 32 is the widest.", width; |
| 1204 | Invalid, Input, Decode)); |
| 1205 | } |
| 1206 | for _ in 0..count { |
| 1207 | entries.push(res!(b.u(width)) as u64 + 1); |
| 1208 | } |
| 1209 | } |
| 1210 | // Under wavefront coding there is one piece per row of coding tree blocks, so the count |
| 1211 | // has to agree with the picture's own geometry -- and the geometry came out of the |
| 1212 | // sequence parameter set, a different NAL unit written at a different time. A slice header |
| 1213 | // read one bit out of step produces a count that is nonsense against it, which makes this |
| 1214 | // the cheapest check there is on the whole header: it is what caught the reading that |
| 1215 | // refused every photograph in the corpus rather than reading its entry points. |
| 1216 | // |
| 1217 | // A segment covering part of a picture names fewer pieces than the picture has rows, and |
| 1218 | // how many fewer is not knowable from the header alone -- so what is checked here is that |
| 1219 | // it names no more, and the exact form is checked by [`whole_picture_rows`] once the number |
| 1220 | // of segments is known. |
| 1221 | if pps.wavefront && !pps.tiles { |
| 1222 | let rows = ((sps.coded_h + sps.ctb_size - 1) / sps.ctb_size) as usize; |
| 1223 | let over = entries.len() + 1 > rows; |
| 1224 | if over { |
| 1225 | return Err(err!( |
| 1226 | "A slice names {} pieces and the picture is {} rows of coding tree blocks \ |
| 1227 | deep. The header has been read out of step.", entries.len() + 1, rows; |
| 1228 | Invalid, Input, Decode)); |
| 1229 | } |
| 1230 | } |
| 1231 | } |
| 1232 | // The header ends with a stop bit and however many zeroes reach the byte boundary, and the |
| 1233 | // arithmetic decoder starts at that boundary (§9.3.1). |
| 1234 | let bits = res!(b.consumed()); |
| 1235 | let data_at = (bits + 1 + 7) / 8; |
| 1236 | if data_at >= body.len() { |
| 1237 | return Err(err!( |
| 1238 | "A slice header of {} bits leaves no data in a payload of {} bytes.", bits, body.len(); |
| 1239 | Invalid, Input, Decode)); |
| 1240 | } |
| 1241 | Ok(Slice { |
| 1242 | first, |
| 1243 | address, |
| 1244 | across_slices, |
| 1245 | pps_id: pps_id as u8, |
| 1246 | kind: slice_kind as u8, |
| 1247 | qp, |
| 1248 | cb_qp_offset: cb_offset, |
| 1249 | cr_qp_offset: cr_offset, |
| 1250 | sao_luma, |
| 1251 | sao_chroma, |
| 1252 | data_at, |
| 1253 | deblocking, |
| 1254 | entries, |
| 1255 | }) |
| 1256 | } |
| 1257 | |
| 1258 | #[cfg(test)] |
| 1259 | mod tests { |
| 1260 | use super::*; |
| 1261 | |
| 1262 | #[test] |
| 1263 | fn test_emulation_prevention_is_undone_00() -> Outcome<()> { |
| 1264 | // A 0x03 after two zeroes is not payload; one after a single zero is. |
| 1265 | req!(rbsp(&[0, 0, 3, 1, 0, 3, 2]), vec![0u8, 0, 1, 0, 3, 2]); |
| 1266 | Ok(()) |
| 1267 | } |
| 1268 | |
| 1269 | #[test] |
| 1270 | fn test_a_truncated_nal_unit_is_refused_01() -> Outcome<()> { |
| 1271 | // Four bytes of length saying eight, with three following. |
| 1272 | let stream = [0u8, 0, 0, 8, 0x26, 1, 9]; |
| 1273 | req!(split_lengthed(&stream, 4).is_err(), true, |
| 1274 | "A unit running past the end of the buffer was read as if it fitted."); |
| 1275 | Ok(()) |
| 1276 | } |
| 1277 | |
| 1278 | |
| 1279 | |
| 1280 | |
| 1281 | |
| 1282 | |
| 1283 | |
| 1284 | #[test] |
| 1285 | fn test_exp_golomb_reads_the_codes_the_specification_names_02() -> Outcome<()> { |
| 1286 | // 1 -> 0, 010 -> 1, 011 -> 2, 00100 -> 3, and the signed mapping 0, 1, -1, 2, -2. |
| 1287 | let mut b = Bits::new(&[0b1010_0110, 0b0100_0000]); |
| 1288 | req!(res!(b.ue()), 0); |
| 1289 | req!(res!(b.ue()), 1); |
| 1290 | req!(res!(b.ue()), 2); |
| 1291 | req!(res!(b.ue()), 3); |
| 1292 | let mut c = Bits::new(&[0b1010_0110, 0b0100_0000]); |
| 1293 | req!(res!(c.se()), 0); |
| 1294 | req!(res!(c.se()), 1); |
| 1295 | req!(res!(c.se()), -1); |
| 1296 | req!(res!(c.se()), 2); |
| 1297 | Ok(()) |
| 1298 | } |
| 1299 | } |
| 1300 | |
| 1301 | // ---------------------------------------------------------------- the whole of one picture |
| 1302 | |
| 1303 | /// Decodes one coded picture: an HEIC tile, or a whole photograph that was not cut into tiles. |
| 1304 | /// |
| 1305 | /// `config` is the `hvcC` record from the container, which carries the parameter sets, and `data` |
| 1306 | /// is the item's bytes -- NAL units with a length prefix each, which is how a HEIF file stores |
| 1307 | /// them rather than with start codes. |
| 1308 | /// |
| 1309 | /// The picture comes back in 4:2:0 at whatever depth it was coded, which for every photograph this |
| 1310 | /// was written against is eight bits. It has **not** been through the deblocking filter or the |
| 1311 | /// sample adaptive offset, which are separate passes over a finished picture. |
| 1312 | pub fn picture(record: &[u8], data: &[u8]) -> Outcome<decode::Picture> { |
| 1313 | let (pic, _sps) = res!(coded(record, data)); |
| 1314 | Ok(pic) |
| 1315 | } |
| 1316 | |
| 1317 | /// The same picture, cropped to the size it is meant to be **shown** at. |
| 1318 | /// |
| 1319 | /// A coded picture is a whole number of coding tree blocks and a shown one is not: a 1920 by 1080 |
| 1320 | /// film is coded 1920 by 1088, and the sequence parameter set's conformance window says which of |
| 1321 | /// those rows are the picture. [`picture`] hands back what was coded, because the HEIC path crops to |
| 1322 | /// the size the container declares instead and cropping twice would take the same rows off again. |
| 1323 | /// A caller with no container to ask -- a film's first frame -- wants this one. |
| 1324 | pub fn picture_shown(record: &[u8], data: &[u8]) -> Outcome<decode::Picture> { |
| 1325 | let (pic, sps) = res!(coded(record, data)); |
| 1326 | let (w, h) = (sps.width as usize, sps.height as usize); |
| 1327 | let (x0, y0) = (sps.show_x0 as usize, sps.show_y0 as usize); |
| 1328 | if x0 == 0 && y0 == 0 && w >= pic.y.w && h >= pic.y.h { |
| 1329 | return Ok(pic); |
| 1330 | } |
| 1331 | Ok(pic.window(x0, y0, w.min(pic.y.w), h.min(pic.y.h))) |
| 1332 | } |
| 1333 | |
| 1334 | /// Decodes one coded picture, and answers the sequence parameter set it was coded against. |
| 1335 | /// |
| 1336 | /// The set is handed back because what a caller does with the picture next depends on it: the |
| 1337 | /// conformance window is in it, and so is everything a caller would otherwise have to parse the |
| 1338 | /// parameter sets again to learn. |
| 1339 | fn coded(record: &[u8], data: &[u8]) -> Outcome<(decode::Picture, Sps)> { |
| 1340 | let cfg = res!(config(record)); |
| 1341 | // Every set the record carries, not the last of each. A photograph out of a |
| 1342 | // camera carries one apiece and either would do; a film carries several, and |
| 1343 | // a slice names which one it was coded against. Keeping the last read meant |
| 1344 | // four films in ten were refused for referring to a set that was in hand all |
| 1345 | // along. |
| 1346 | let mut seqs: Vec<Sps> = Vec::new(); |
| 1347 | let mut pics: Vec<Pps> = Vec::new(); |
| 1348 | for unit in &cfg.sets { |
| 1349 | match unit.kind { |
| 1350 | nal::SPS => seqs.push(res!(sps(&unit.body))), |
| 1351 | nal::PPS => pics.push(res!(pps(&unit.body))), |
| 1352 | _ => {}, |
| 1353 | } |
| 1354 | } |
| 1355 | if seqs.is_empty() { |
| 1356 | return Err(err!( |
| 1357 | "The decoder configuration carries no sequence parameter set."; Invalid, Input)); |
| 1358 | } |
| 1359 | if pics.is_empty() { |
| 1360 | return Err(err!( |
| 1361 | "The decoder configuration carries no picture parameter set."; Invalid, Input)); |
| 1362 | } |
| 1363 | |
| 1364 | // And the slices are in the item's own bytes. **Every** slice of the picture, not the first: |
| 1365 | // a photograph is one slice and a film's frame need not be, and a picture read from one of |
| 1366 | // four segments is a quarter of a picture. |
| 1367 | let units = res!(split_lengthed(data, cfg.length_size)); |
| 1368 | let mut heads: Vec<(Slice, usize)> = Vec::new(); |
| 1369 | let mut chosen: Option<(Sps, Pps)> = None; |
| 1370 | for (i, unit) in units.iter().enumerate() { |
| 1371 | match unit.kind { |
| 1372 | nal::IDR_W_RADL | nal::IDR_N_LP | 21 => { |
| 1373 | // Which sets this slice was coded against: the picture set it |
| 1374 | // names, and the sequence set that one belongs to. |
| 1375 | let want = res!(slice_pps_id(&unit.body)); |
| 1376 | let pps = match pics.iter().find(|p| p.id == want) { |
| 1377 | Some(p) => p.clone(), |
| 1378 | None => return Err(err!( |
| 1379 | "A slice references picture parameter set {}, and the configuration \ |
| 1380 | carries {}.", want, |
| 1381 | pics.iter().map(|p| p.id.to_string()).collect::<Vec<_>>().join(", "); |
| 1382 | Invalid, Input, Missing)), |
| 1383 | }; |
| 1384 | let sps = match seqs.iter().find(|s| s.id == pps.sps_id) { |
| 1385 | Some(s) => s.clone(), |
| 1386 | None => return Err(err!( |
| 1387 | "Picture parameter set {} belongs to sequence parameter set {}, and the \ |
| 1388 | configuration carries {}.", pps.id, pps.sps_id, |
| 1389 | seqs.iter().map(|s| s.id.to_string()).collect::<Vec<_>>().join(", "); |
| 1390 | Invalid, Input, Missing)), |
| 1391 | }; |
| 1392 | let head = res!(slice_of(unit.kind, &unit.body, &sps, &pps)); |
| 1393 | // A second coded picture in the same access unit is somebody else's frame: this |
| 1394 | // reads the first picture, and the first picture ends where the next one begins. |
| 1395 | if head.first && !heads.is_empty() { |
| 1396 | break; |
| 1397 | } |
| 1398 | match &chosen { |
| 1399 | Some((have_sps, have_pps)) => { |
| 1400 | if have_sps.id != sps.id || have_pps.id != pps.id { |
| 1401 | return Err(err!( |
| 1402 | "Two slices of one picture reference different parameter sets."; |
| 1403 | Invalid, Input, Mismatch)); |
| 1404 | } |
| 1405 | }, |
| 1406 | None => chosen = Some((sps, pps)), |
| 1407 | } |
| 1408 | heads.push((head, i)); |
| 1409 | }, |
| 1410 | _ => {}, |
| 1411 | } |
| 1412 | } |
| 1413 | let (sps, pps) = match chosen { |
| 1414 | Some(pair) => pair, |
| 1415 | None => return Err(err!("Those bytes hold no coded slice."; Invalid, Input, Decode)), |
| 1416 | }; |
| 1417 | // A picture that is one slice must name one piece a row of blocks, since that is what |
| 1418 | // wavefront coding is. It is the cheapest check there is on the whole header -- the count and |
| 1419 | // the geometry come out of different NAL units written at different times -- and it is what |
| 1420 | // caught the reading that refused every photograph in the corpus. |
| 1421 | if heads.len() == 1 && pps.wavefront && !pps.tiles { |
| 1422 | let rows = ((sps.coded_h + sps.ctb_size - 1) / sps.ctb_size) as usize; |
| 1423 | let named = heads[0].0.entries.len() + 1; |
| 1424 | if named != rows { |
| 1425 | return Err(err!( |
| 1426 | "A slice names {} pieces and the picture is {} rows of coding tree blocks deep. \ |
| 1427 | The header has been read out of step.", named, rows; |
| 1428 | Invalid, Input, Decode)); |
| 1429 | } |
| 1430 | } |
| 1431 | // The header was read from the unescaped payload; the data after it has to be handed over |
| 1432 | // escaped, because that is what the entry point offsets count. |
| 1433 | let parts: Vec<(&Slice, &[u8])> = heads.iter() |
| 1434 | .map(|(head, i)| { |
| 1435 | let raw = &units[*i].raw; |
| 1436 | (head, &raw[escaped_at(raw, head.data_at).min(raw.len())..]) |
| 1437 | }) |
| 1438 | .collect(); |
| 1439 | let pic = res!(decode::picture_of(&sps, &pps, &parts)); |
| 1440 | Ok((pic, sps)) |
| 1441 | } |