git.lucas.co / go_mono
git clone https://git.lucas.co/go_mono.git

vp8/decode.go (12K)

  1 // Copyright 2011 The Go Authors. All rights reserved.
  2 // Use of this source code is governed by a BSD-style
  3 // license that can be found in the LICENSE file.
  4 
  5 // Package vp8 implements a decoder for the VP8 lossy image format.
  6 //
  7 // The VP8 specification is RFC 6386.
  8 package vp8 // import "golang.org/x/image/vp8"
  9 
 10 // This file implements the top-level decoding algorithm.
 11 
 12 import (
 13 	"errors"
 14 	"image"
 15 	"io"
 16 )
 17 
 18 // limitReader wraps an io.Reader to read at most n bytes from it.
 19 type limitReader struct {
 20 	r io.Reader
 21 	n int
 22 }
 23 
 24 // ReadFull reads exactly len(p) bytes into p.
 25 func (r *limitReader) ReadFull(p []byte) error {
 26 	if len(p) > r.n {
 27 		return io.ErrUnexpectedEOF
 28 	}
 29 	n, err := io.ReadFull(r.r, p)
 30 	r.n -= n
 31 	return err
 32 }
 33 
 34 // FrameHeader is a frame header, as specified in section 9.1.
 35 type FrameHeader struct {
 36 	KeyFrame          bool
 37 	VersionNumber     uint8
 38 	ShowFrame         bool
 39 	FirstPartitionLen uint32
 40 	Width             int
 41 	Height            int
 42 	XScale            uint8
 43 	YScale            uint8
 44 }
 45 
 46 const (
 47 	nSegment     = 4
 48 	nSegmentProb = 3
 49 )
 50 
 51 // segmentHeader holds segment-related header information.
 52 type segmentHeader struct {
 53 	useSegment     bool
 54 	updateMap      bool
 55 	relativeDelta  bool
 56 	quantizer      [nSegment]int8
 57 	filterStrength [nSegment]int8
 58 	prob           [nSegmentProb]uint8
 59 }
 60 
 61 const (
 62 	nRefLFDelta  = 4
 63 	nModeLFDelta = 4
 64 )
 65 
 66 // filterHeader holds filter-related header information.
 67 type filterHeader struct {
 68 	simple          bool
 69 	level           int8
 70 	sharpness       uint8
 71 	useLFDelta      bool
 72 	refLFDelta      [nRefLFDelta]int8
 73 	modeLFDelta     [nModeLFDelta]int8
 74 	perSegmentLevel [nSegment]int8
 75 }
 76 
 77 // mb is the per-macroblock decode state. A decoder maintains mbw+1 of these
 78 // as it is decoding macroblocks left-to-right and top-to-bottom: mbw for the
 79 // macroblocks in the row above, and one for the macroblock to the left.
 80 type mb struct {
 81 	// pred is the predictor mode for the 4 bottom or right 4x4 luma regions.
 82 	pred [4]uint8
 83 	// nzMask is a mask of 8 bits: 4 for the bottom or right 4x4 luma regions,
 84 	// and 2 + 2 for the bottom or right 4x4 chroma regions. A 1 bit indicates
 85 	// that region has non-zero coefficients.
 86 	nzMask uint8
 87 	// nzY16 is a 0/1 value that is 1 if the macroblock used Y16 prediction and
 88 	// had non-zero coefficients.
 89 	nzY16 uint8
 90 }
 91 
 92 // Decoder decodes VP8 bitstreams into frames. Decoding one frame consists of
 93 // calling Init, DecodeFrameHeader and then DecodeFrame in that order.
 94 // A Decoder can be re-used to decode multiple frames.
 95 type Decoder struct {
 96 	// r is the input bitsream.
 97 	r limitReader
 98 	// scratch is a scratch buffer.
 99 	scratch [8]byte
100 	// img is the YCbCr image to decode into.
101 	img *image.YCbCr
102 	// mbw and mbh are the number of 16x16 macroblocks wide and high the image is.
103 	mbw, mbh int
104 	// frameHeader is the frame header. When decoding multiple frames,
105 	// frames that aren't key frames will inherit the Width, Height,
106 	// XScale and YScale of the most recent key frame.
107 	frameHeader FrameHeader
108 	// Other headers.
109 	segmentHeader segmentHeader
110 	filterHeader  filterHeader
111 	// The image data is divided into a number of independent partitions.
112 	// There is 1 "first partition" and between 1 and 8 "other partitions"
113 	// for coefficient data.
114 	fp  partition
115 	op  [8]partition
116 	nOP int
117 	// Quantization factors.
118 	quant [nSegment]quant
119 	// DCT/WHT coefficient decoding probabilities.
120 	tokenProb   [nPlane][nBand][nContext][nProb]uint8
121 	useSkipProb bool
122 	skipProb    uint8
123 	// Loop filter parameters.
124 	filterParams      [nSegment][2]filterParam
125 	perMBFilterParams []filterParam
126 
127 	// The eight fields below relate to the current macroblock being decoded.
128 	//
129 	// Segment-based adjustments.
130 	segment int
131 	// Per-macroblock state for the macroblock immediately left of and those
132 	// macroblocks immediately above the current macroblock.
133 	leftMB mb
134 	upMB   []mb
135 	// Bitmasks for which 4x4 regions of coeff contain non-zero coefficients.
136 	nzDCMask, nzACMask uint32
137 	// Predictor modes.
138 	usePredY16 bool // The libwebp C code calls this !is_i4x4_.
139 	predY16    uint8
140 	predC8     uint8
141 	predY4     [4][4]uint8
142 
143 	// The two fields below form a workspace for reconstructing a macroblock.
144 	// Their specific sizes are documented in reconstruct.go.
145 	coeff [1*16*16 + 2*8*8 + 1*4*4]int16
146 	ybr   [1 + 16 + 1 + 8][32]uint8
147 }
148 
149 // NewDecoder returns a new Decoder.
150 func NewDecoder() *Decoder {
151 	return &Decoder{}
152 }
153 
154 // Init initializes the decoder to read at most n bytes from r.
155 func (d *Decoder) Init(r io.Reader, n int) {
156 	d.r = limitReader{r, n}
157 }
158 
159 // DecodeFrameHeader decodes the frame header.
160 func (d *Decoder) DecodeFrameHeader() (fh FrameHeader, err error) {
161 	// All frame headers are at least 3 bytes long.
162 	b := d.scratch[:3]
163 	if err = d.r.ReadFull(b); err != nil {
164 		return
165 	}
166 	d.frameHeader.KeyFrame = (b[0] & 1) == 0
167 	d.frameHeader.VersionNumber = (b[0] >> 1) & 7
168 	d.frameHeader.ShowFrame = (b[0]>>4)&1 == 1
169 	d.frameHeader.FirstPartitionLen = uint32(b[0])>>5 | uint32(b[1])<<3 | uint32(b[2])<<11
170 	if !d.frameHeader.KeyFrame {
171 		return d.frameHeader, nil
172 	}
173 	// Frame headers for key frames are an additional 7 bytes long.
174 	b = d.scratch[:7]
175 	if err = d.r.ReadFull(b); err != nil {
176 		return
177 	}
178 	// Check the magic sync code.
179 	if b[0] != 0x9d || b[1] != 0x01 || b[2] != 0x2a {
180 		err = errors.New("vp8: invalid format")
181 		return
182 	}
183 	d.frameHeader.Width = int(b[4]&0x3f)<<8 | int(b[3])
184 	d.frameHeader.Height = int(b[6]&0x3f)<<8 | int(b[5])
185 	d.frameHeader.XScale = b[4] >> 6
186 	d.frameHeader.YScale = b[6] >> 6
187 	d.mbw = (d.frameHeader.Width + 0x0f) >> 4
188 	d.mbh = (d.frameHeader.Height + 0x0f) >> 4
189 	d.segmentHeader = segmentHeader{
190 		prob: [3]uint8{0xff, 0xff, 0xff},
191 	}
192 	d.tokenProb = defaultTokenProb
193 	d.segment = 0
194 	return d.frameHeader, nil
195 }
196 
197 // ensureImg ensures that d.img is large enough to hold the decoded frame.
198 func (d *Decoder) ensureImg() {
199 	if d.img != nil {
200 		p0, p1 := d.img.Rect.Min, d.img.Rect.Max
201 		if p0.X == 0 && p0.Y == 0 && p1.X >= 16*d.mbw && p1.Y >= 16*d.mbh {
202 			return
203 		}
204 	}
205 	m := image.NewYCbCr(image.Rect(0, 0, 16*d.mbw, 16*d.mbh), image.YCbCrSubsampleRatio420)
206 	d.img = m.SubImage(image.Rect(0, 0, d.frameHeader.Width, d.frameHeader.Height)).(*image.YCbCr)
207 	d.perMBFilterParams = make([]filterParam, d.mbw*d.mbh)
208 	d.upMB = make([]mb, d.mbw)
209 }
210 
211 // parseSegmentHeader parses the segment header, as specified in section 9.3.
212 func (d *Decoder) parseSegmentHeader() {
213 	d.segmentHeader.useSegment = d.fp.readBit(uniformProb)
214 	if !d.segmentHeader.useSegment {
215 		d.segmentHeader.updateMap = false
216 		return
217 	}
218 	d.segmentHeader.updateMap = d.fp.readBit(uniformProb)
219 	if d.fp.readBit(uniformProb) {
220 		d.segmentHeader.relativeDelta = !d.fp.readBit(uniformProb)
221 		for i := range d.segmentHeader.quantizer {
222 			d.segmentHeader.quantizer[i] = int8(d.fp.readOptionalInt(uniformProb, 7))
223 		}
224 		for i := range d.segmentHeader.filterStrength {
225 			d.segmentHeader.filterStrength[i] = int8(d.fp.readOptionalInt(uniformProb, 6))
226 		}
227 	}
228 	if !d.segmentHeader.updateMap {
229 		return
230 	}
231 	for i := range d.segmentHeader.prob {
232 		if d.fp.readBit(uniformProb) {
233 			d.segmentHeader.prob[i] = uint8(d.fp.readUint(uniformProb, 8))
234 		} else {
235 			d.segmentHeader.prob[i] = 0xff
236 		}
237 	}
238 }
239 
240 // parseFilterHeader parses the filter header, as specified in section 9.4.
241 func (d *Decoder) parseFilterHeader() {
242 	d.filterHeader.simple = d.fp.readBit(uniformProb)
243 	d.filterHeader.level = int8(d.fp.readUint(uniformProb, 6))
244 	d.filterHeader.sharpness = uint8(d.fp.readUint(uniformProb, 3))
245 	d.filterHeader.useLFDelta = d.fp.readBit(uniformProb)
246 	if d.filterHeader.useLFDelta && d.fp.readBit(uniformProb) {
247 		for i := range d.filterHeader.refLFDelta {
248 			d.filterHeader.refLFDelta[i] = int8(d.fp.readOptionalInt(uniformProb, 6))
249 		}
250 		for i := range d.filterHeader.modeLFDelta {
251 			d.filterHeader.modeLFDelta[i] = int8(d.fp.readOptionalInt(uniformProb, 6))
252 		}
253 	}
254 	if d.filterHeader.level == 0 {
255 		return
256 	}
257 	if d.segmentHeader.useSegment {
258 		for i := range d.filterHeader.perSegmentLevel {
259 			strength := d.segmentHeader.filterStrength[i]
260 			if d.segmentHeader.relativeDelta {
261 				strength += d.filterHeader.level
262 			}
263 			d.filterHeader.perSegmentLevel[i] = strength
264 		}
265 	} else {
266 		d.filterHeader.perSegmentLevel[0] = d.filterHeader.level
267 	}
268 	d.computeFilterParams()
269 }
270 
271 // parseOtherPartitions parses the other partitions, as specified in section 9.5.
272 func (d *Decoder) parseOtherPartitions() error {
273 	const maxNOP = 1 << 3
274 	var partLens [maxNOP]int
275 	d.nOP = 1 << d.fp.readUint(uniformProb, 2)
276 
277 	// The final partition length is implied by the remaining chunk data
278 	// (d.r.n) and the other d.nOP-1 partition lengths. Those d.nOP-1 partition
279 	// lengths are stored as 24-bit uints, i.e. up to 16 MiB per partition.
280 	n := 3 * (d.nOP - 1)
281 	partLens[d.nOP-1] = d.r.n - n
282 	if partLens[d.nOP-1] < 0 {
283 		return io.ErrUnexpectedEOF
284 	}
285 	if n > 0 {
286 		buf := make([]byte, n)
287 		if err := d.r.ReadFull(buf); err != nil {
288 			return err
289 		}
290 		for i := 0; i < d.nOP-1; i++ {
291 			pl := int(buf[3*i+0]) | int(buf[3*i+1])<<8 | int(buf[3*i+2])<<16
292 			if pl > partLens[d.nOP-1] {
293 				return io.ErrUnexpectedEOF
294 			}
295 			partLens[i] = pl
296 			partLens[d.nOP-1] -= pl
297 		}
298 	}
299 
300 	// We check if the final partition length can also fit into a 24-bit uint.
301 	// Strictly speaking, this isn't part of the spec, but it guards against a
302 	// malicious WEBP image that is too large to ReadFull the encoded DCT
303 	// coefficients into memory, whether that's because the actual WEBP file is
304 	// too large, or whether its RIFF metadata lists too large a chunk.
305 	if 1<<24 <= partLens[d.nOP-1] {
306 		return errors.New("vp8: too much data to decode")
307 	}
308 
309 	buf := make([]byte, d.r.n)
310 	if err := d.r.ReadFull(buf); err != nil {
311 		return err
312 	}
313 	for i, pl := range partLens {
314 		if i == d.nOP {
315 			break
316 		}
317 		d.op[i].init(buf[:pl])
318 		buf = buf[pl:]
319 	}
320 	return nil
321 }
322 
323 // parseOtherHeaders parses header information other than the frame header.
324 func (d *Decoder) parseOtherHeaders() error {
325 	// Initialize and parse the first partition.
326 	firstPartition := make([]byte, d.frameHeader.FirstPartitionLen)
327 	if err := d.r.ReadFull(firstPartition); err != nil {
328 		return err
329 	}
330 	d.fp.init(firstPartition)
331 	if d.frameHeader.KeyFrame {
332 		// Read and ignore the color space and pixel clamp values. They are
333 		// specified in section 9.2, but are unimplemented.
334 		d.fp.readBit(uniformProb)
335 		d.fp.readBit(uniformProb)
336 	}
337 	d.parseSegmentHeader()
338 	d.parseFilterHeader()
339 	if err := d.parseOtherPartitions(); err != nil {
340 		return err
341 	}
342 	d.parseQuant()
343 	if !d.frameHeader.KeyFrame {
344 		// Golden and AltRef frames are specified in section 9.7.
345 		// TODO(nigeltao): implement. Note that they are only used for video, not still images.
346 		return errors.New("vp8: Golden / AltRef frames are not implemented")
347 	}
348 	// Read and ignore the refreshLastFrameBuffer bit, specified in section 9.8.
349 	// It applies only to video, and not still images.
350 	d.fp.readBit(uniformProb)
351 	d.parseTokenProb()
352 	d.useSkipProb = d.fp.readBit(uniformProb)
353 	if d.useSkipProb {
354 		d.skipProb = uint8(d.fp.readUint(uniformProb, 8))
355 	}
356 	if d.fp.unexpectedEOF {
357 		return io.ErrUnexpectedEOF
358 	}
359 	return nil
360 }
361 
362 // DecodeFrame decodes the frame and returns it as an YCbCr image.
363 // The image's contents are valid up until the next call to Decoder.Init.
364 func (d *Decoder) DecodeFrame() (*image.YCbCr, error) {
365 	d.ensureImg()
366 	if err := d.parseOtherHeaders(); err != nil {
367 		return nil, err
368 	}
369 	// Reconstruct the rows.
370 	for mbx := 0; mbx < d.mbw; mbx++ {
371 		d.upMB[mbx] = mb{}
372 	}
373 	for mby := 0; mby < d.mbh; mby++ {
374 		d.leftMB = mb{}
375 		for mbx := 0; mbx < d.mbw; mbx++ {
376 			skip := d.reconstruct(mbx, mby)
377 			fs := d.filterParams[d.segment][btou(!d.usePredY16)]
378 			fs.inner = fs.inner || !skip
379 			d.perMBFilterParams[d.mbw*mby+mbx] = fs
380 		}
381 	}
382 	if d.fp.unexpectedEOF {
383 		return nil, io.ErrUnexpectedEOF
384 	}
385 	for i := 0; i < d.nOP; i++ {
386 		if d.op[i].unexpectedEOF {
387 			return nil, io.ErrUnexpectedEOF
388 		}
389 	}
390 	// Apply the loop filter.
391 	//
392 	// Even if we are using per-segment levels, section 15 says that "loop
393 	// filtering must be skipped entirely if loop_filter_level at either the
394 	// frame header level or macroblock override level is 0".
395 	if d.filterHeader.level != 0 {
396 		if d.filterHeader.simple {
397 			d.simpleFilter()
398 		} else {
399 			d.normalFilter()
400 		}
401 	}
402 	return d.img, nil
403 }