...
Run Format

Source file src/image/jpeg/reader.go

  // Copyright 2009 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  // Package jpeg implements a JPEG image decoder and encoder.
  //
  // JPEG is defined in ITU-T T.81: http://www.w3.org/Graphics/JPEG/itu-t81.pdf.
  package jpeg
  
  import (
  	"image"
  	"image/color"
  	"image/internal/imageutil"
  	"io"
  )
  
  // TODO(nigeltao): fix up the doc comment style so that sentences start with
  // the name of the type or function that they annotate.
  
  // A FormatError reports that the input is not a valid JPEG.
  type FormatError string
  
  func (e FormatError) Error() string { return "invalid JPEG format: " + string(e) }
  
  // An UnsupportedError reports that the input uses a valid but unimplemented JPEG feature.
  type UnsupportedError string
  
  func (e UnsupportedError) Error() string { return "unsupported JPEG feature: " + string(e) }
  
  var errUnsupportedSubsamplingRatio = UnsupportedError("luma/chroma subsampling ratio")
  
  // Component specification, specified in section B.2.2.
  type component struct {
  	h  int   // Horizontal sampling factor.
  	v  int   // Vertical sampling factor.
  	c  uint8 // Component identifier.
  	tq uint8 // Quantization table destination selector.
  }
  
  const (
  	dcTable = 0
  	acTable = 1
  	maxTc   = 1
  	maxTh   = 3
  	maxTq   = 3
  
  	maxComponents = 4
  )
  
  const (
  	sof0Marker = 0xc0 // Start Of Frame (Baseline).
  	sof1Marker = 0xc1 // Start Of Frame (Extended Sequential).
  	sof2Marker = 0xc2 // Start Of Frame (Progressive).
  	dhtMarker  = 0xc4 // Define Huffman Table.
  	rst0Marker = 0xd0 // ReSTart (0).
  	rst7Marker = 0xd7 // ReSTart (7).
  	soiMarker  = 0xd8 // Start Of Image.
  	eoiMarker  = 0xd9 // End Of Image.
  	sosMarker  = 0xda // Start Of Scan.
  	dqtMarker  = 0xdb // Define Quantization Table.
  	driMarker  = 0xdd // Define Restart Interval.
  	comMarker  = 0xfe // COMment.
  	// "APPlication specific" markers aren't part of the JPEG spec per se,
  	// but in practice, their use is described at
  	// http://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/JPEG.html
  	app0Marker  = 0xe0
  	app14Marker = 0xee
  	app15Marker = 0xef
  )
  
  // See http://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/JPEG.html#Adobe
  const (
  	adobeTransformUnknown = 0
  	adobeTransformYCbCr   = 1
  	adobeTransformYCbCrK  = 2
  )
  
  // unzig maps from the zig-zag ordering to the natural ordering. For example,
  // unzig[3] is the column and row of the fourth element in zig-zag order. The
  // value is 16, which means first column (16%8 == 0) and third row (16/8 == 2).
  var unzig = [blockSize]int{
  	0, 1, 8, 16, 9, 2, 3, 10,
  	17, 24, 32, 25, 18, 11, 4, 5,
  	12, 19, 26, 33, 40, 48, 41, 34,
  	27, 20, 13, 6, 7, 14, 21, 28,
  	35, 42, 49, 56, 57, 50, 43, 36,
  	29, 22, 15, 23, 30, 37, 44, 51,
  	58, 59, 52, 45, 38, 31, 39, 46,
  	53, 60, 61, 54, 47, 55, 62, 63,
  }
  
  // Deprecated: Reader is deprecated.
  type Reader interface {
  	io.ByteReader
  	io.Reader
  }
  
  // bits holds the unprocessed bits that have been taken from the byte-stream.
  // The n least significant bits of a form the unread bits, to be read in MSB to
  // LSB order.
  type bits struct {
  	a uint32 // accumulator.
  	m uint32 // mask. m==1<<(n-1) when n>0, with m==0 when n==0.
  	n int32  // the number of unread bits in a.
  }
  
  type decoder struct {
  	r    io.Reader
  	bits bits
  	// bytes is a byte buffer, similar to a bufio.Reader, except that it
  	// has to be able to unread more than 1 byte, due to byte stuffing.
  	// Byte stuffing is specified in section F.1.2.3.
  	bytes struct {
  		// buf[i:j] are the buffered bytes read from the underlying
  		// io.Reader that haven't yet been passed further on.
  		buf  [4096]byte
  		i, j int
  		// nUnreadable is the number of bytes to back up i after
  		// overshooting. It can be 0, 1 or 2.
  		nUnreadable int
  	}
  	width, height int
  
  	img1        *image.Gray
  	img3        *image.YCbCr
  	blackPix    []byte
  	blackStride int
  
  	ri                  int // Restart Interval.
  	nComp               int
  	progressive         bool
  	jfif                bool
  	adobeTransformValid bool
  	adobeTransform      uint8
  	eobRun              uint16 // End-of-Band run, specified in section G.1.2.2.
  
  	comp       [maxComponents]component
  	progCoeffs [maxComponents][]block // Saved state between progressive-mode scans.
  	huff       [maxTc + 1][maxTh + 1]huffman
  	quant      [maxTq + 1]block // Quantization tables, in zig-zag order.
  	tmp        [2 * blockSize]byte
  }
  
  // fill fills up the d.bytes.buf buffer from the underlying io.Reader. It
  // should only be called when there are no unread bytes in d.bytes.
  func (d *decoder) fill() error {
  	if d.bytes.i != d.bytes.j {
  		panic("jpeg: fill called when unread bytes exist")
  	}
  	// Move the last 2 bytes to the start of the buffer, in case we need
  	// to call unreadByteStuffedByte.
  	if d.bytes.j > 2 {
  		d.bytes.buf[0] = d.bytes.buf[d.bytes.j-2]
  		d.bytes.buf[1] = d.bytes.buf[d.bytes.j-1]
  		d.bytes.i, d.bytes.j = 2, 2
  	}
  	// Fill in the rest of the buffer.
  	n, err := d.r.Read(d.bytes.buf[d.bytes.j:])
  	d.bytes.j += n
  	if n > 0 {
  		err = nil
  	}
  	return err
  }
  
  // unreadByteStuffedByte undoes the most recent readByteStuffedByte call,
  // giving a byte of data back from d.bits to d.bytes. The Huffman look-up table
  // requires at least 8 bits for look-up, which means that Huffman decoding can
  // sometimes overshoot and read one or two too many bytes. Two-byte overshoot
  // can happen when expecting to read a 0xff 0x00 byte-stuffed byte.
  func (d *decoder) unreadByteStuffedByte() {
  	d.bytes.i -= d.bytes.nUnreadable
  	d.bytes.nUnreadable = 0
  	if d.bits.n >= 8 {
  		d.bits.a >>= 8
  		d.bits.n -= 8
  		d.bits.m >>= 8
  	}
  }
  
  // readByte returns the next byte, whether buffered or not buffered. It does
  // not care about byte stuffing.
  func (d *decoder) readByte() (x byte, err error) {
  	for d.bytes.i == d.bytes.j {
  		if err = d.fill(); err != nil {
  			return 0, err
  		}
  	}
  	x = d.bytes.buf[d.bytes.i]
  	d.bytes.i++
  	d.bytes.nUnreadable = 0
  	return x, nil
  }
  
  // errMissingFF00 means that readByteStuffedByte encountered an 0xff byte (a
  // marker byte) that wasn't the expected byte-stuffed sequence 0xff, 0x00.
  var errMissingFF00 = FormatError("missing 0xff00 sequence")
  
  // readByteStuffedByte is like readByte but is for byte-stuffed Huffman data.
  func (d *decoder) readByteStuffedByte() (x byte, err error) {
  	// Take the fast path if d.bytes.buf contains at least two bytes.
  	if d.bytes.i+2 <= d.bytes.j {
  		x = d.bytes.buf[d.bytes.i]
  		d.bytes.i++
  		d.bytes.nUnreadable = 1
  		if x != 0xff {
  			return x, err
  		}
  		if d.bytes.buf[d.bytes.i] != 0x00 {
  			return 0, errMissingFF00
  		}
  		d.bytes.i++
  		d.bytes.nUnreadable = 2
  		return 0xff, nil
  	}
  
  	d.bytes.nUnreadable = 0
  
  	x, err = d.readByte()
  	if err != nil {
  		return 0, err
  	}
  	d.bytes.nUnreadable = 1
  	if x != 0xff {
  		return x, nil
  	}
  
  	x, err = d.readByte()
  	if err != nil {
  		return 0, err
  	}
  	d.bytes.nUnreadable = 2
  	if x != 0x00 {
  		return 0, errMissingFF00
  	}
  	return 0xff, nil
  }
  
  // readFull reads exactly len(p) bytes into p. It does not care about byte
  // stuffing.
  func (d *decoder) readFull(p []byte) error {
  	// Unread the overshot bytes, if any.
  	if d.bytes.nUnreadable != 0 {
  		if d.bits.n >= 8 {
  			d.unreadByteStuffedByte()
  		}
  		d.bytes.nUnreadable = 0
  	}
  
  	for {
  		n := copy(p, d.bytes.buf[d.bytes.i:d.bytes.j])
  		p = p[n:]
  		d.bytes.i += n
  		if len(p) == 0 {
  			break
  		}
  		if err := d.fill(); err != nil {
  			if err == io.EOF {
  				err = io.ErrUnexpectedEOF
  			}
  			return err
  		}
  	}
  	return nil
  }
  
  // ignore ignores the next n bytes.
  func (d *decoder) ignore(n int) error {
  	// Unread the overshot bytes, if any.
  	if d.bytes.nUnreadable != 0 {
  		if d.bits.n >= 8 {
  			d.unreadByteStuffedByte()
  		}
  		d.bytes.nUnreadable = 0
  	}
  
  	for {
  		m := d.bytes.j - d.bytes.i
  		if m > n {
  			m = n
  		}
  		d.bytes.i += m
  		n -= m
  		if n == 0 {
  			break
  		}
  		if err := d.fill(); err != nil {
  			if err == io.EOF {
  				err = io.ErrUnexpectedEOF
  			}
  			return err
  		}
  	}
  	return nil
  }
  
  // Specified in section B.2.2.
  func (d *decoder) processSOF(n int) error {
  	if d.nComp != 0 {
  		return FormatError("multiple SOF markers")
  	}
  	switch n {
  	case 6 + 3*1: // Grayscale image.
  		d.nComp = 1
  	case 6 + 3*3: // YCbCr or RGB image.
  		d.nComp = 3
  	case 6 + 3*4: // YCbCrK or CMYK image.
  		d.nComp = 4
  	default:
  		return UnsupportedError("number of components")
  	}
  	if err := d.readFull(d.tmp[:n]); err != nil {
  		return err
  	}
  	// We only support 8-bit precision.
  	if d.tmp[0] != 8 {
  		return UnsupportedError("precision")
  	}
  	d.height = int(d.tmp[1])<<8 + int(d.tmp[2])
  	d.width = int(d.tmp[3])<<8 + int(d.tmp[4])
  	if int(d.tmp[5]) != d.nComp {
  		return FormatError("SOF has wrong length")
  	}
  
  	for i := 0; i < d.nComp; i++ {
  		d.comp[i].c = d.tmp[6+3*i]
  		// Section B.2.2 states that "the value of C_i shall be different from
  		// the values of C_1 through C_(i-1)".
  		for j := 0; j < i; j++ {
  			if d.comp[i].c == d.comp[j].c {
  				return FormatError("repeated component identifier")
  			}
  		}
  
  		d.comp[i].tq = d.tmp[8+3*i]
  		if d.comp[i].tq > maxTq {
  			return FormatError("bad Tq value")
  		}
  
  		hv := d.tmp[7+3*i]
  		h, v := int(hv>>4), int(hv&0x0f)
  		if h < 1 || 4 < h || v < 1 || 4 < v {
  			return FormatError("luma/chroma subsampling ratio")
  		}
  		if h == 3 || v == 3 {
  			return errUnsupportedSubsamplingRatio
  		}
  		switch d.nComp {
  		case 1:
  			// If a JPEG image has only one component, section A.2 says "this data
  			// is non-interleaved by definition" and section A.2.2 says "[in this
  			// case...] the order of data units within a scan shall be left-to-right
  			// and top-to-bottom... regardless of the values of H_1 and V_1". Section
  			// 4.8.2 also says "[for non-interleaved data], the MCU is defined to be
  			// one data unit". Similarly, section A.1.1 explains that it is the ratio
  			// of H_i to max_j(H_j) that matters, and similarly for V. For grayscale
  			// images, H_1 is the maximum H_j for all components j, so that ratio is
  			// always 1. The component's (h, v) is effectively always (1, 1): even if
  			// the nominal (h, v) is (2, 1), a 20x5 image is encoded in three 8x8
  			// MCUs, not two 16x8 MCUs.
  			h, v = 1, 1
  
  		case 3:
  			// For YCbCr images, we only support 4:4:4, 4:4:0, 4:2:2, 4:2:0,
  			// 4:1:1 or 4:1:0 chroma subsampling ratios. This implies that the
  			// (h, v) values for the Y component are either (1, 1), (1, 2),
  			// (2, 1), (2, 2), (4, 1) or (4, 2), and the Y component's values
  			// must be a multiple of the Cb and Cr component's values. We also
  			// assume that the two chroma components have the same subsampling
  			// ratio.
  			switch i {
  			case 0: // Y.
  				// We have already verified, above, that h and v are both
  				// either 1, 2 or 4, so invalid (h, v) combinations are those
  				// with v == 4.
  				if v == 4 {
  					return errUnsupportedSubsamplingRatio
  				}
  			case 1: // Cb.
  				if d.comp[0].h%h != 0 || d.comp[0].v%v != 0 {
  					return errUnsupportedSubsamplingRatio
  				}
  			case 2: // Cr.
  				if d.comp[1].h != h || d.comp[1].v != v {
  					return errUnsupportedSubsamplingRatio
  				}
  			}
  
  		case 4:
  			// For 4-component images (either CMYK or YCbCrK), we only support two
  			// hv vectors: [0x11 0x11 0x11 0x11] and [0x22 0x11 0x11 0x22].
  			// Theoretically, 4-component JPEG images could mix and match hv values
  			// but in practice, those two combinations are the only ones in use,
  			// and it simplifies the applyBlack code below if we can assume that:
  			//	- for CMYK, the C and K channels have full samples, and if the M
  			//	  and Y channels subsample, they subsample both horizontally and
  			//	  vertically.
  			//	- for YCbCrK, the Y and K channels have full samples.
  			switch i {
  			case 0:
  				if hv != 0x11 && hv != 0x22 {
  					return errUnsupportedSubsamplingRatio
  				}
  			case 1, 2:
  				if hv != 0x11 {
  					return errUnsupportedSubsamplingRatio
  				}
  			case 3:
  				if d.comp[0].h != h || d.comp[0].v != v {
  					return errUnsupportedSubsamplingRatio
  				}
  			}
  		}
  
  		d.comp[i].h = h
  		d.comp[i].v = v
  	}
  	return nil
  }
  
  // Specified in section B.2.4.1.
  func (d *decoder) processDQT(n int) error {
  loop:
  	for n > 0 {
  		n--
  		x, err := d.readByte()
  		if err != nil {
  			return err
  		}
  		tq := x & 0x0f
  		if tq > maxTq {
  			return FormatError("bad Tq value")
  		}
  		switch x >> 4 {
  		default:
  			return FormatError("bad Pq value")
  		case 0:
  			if n < blockSize {
  				break loop
  			}
  			n -= blockSize
  			if err := d.readFull(d.tmp[:blockSize]); err != nil {
  				return err
  			}
  			for i := range d.quant[tq] {
  				d.quant[tq][i] = int32(d.tmp[i])
  			}
  		case 1:
  			if n < 2*blockSize {
  				break loop
  			}
  			n -= 2 * blockSize
  			if err := d.readFull(d.tmp[:2*blockSize]); err != nil {
  				return err
  			}
  			for i := range d.quant[tq] {
  				d.quant[tq][i] = int32(d.tmp[2*i])<<8 | int32(d.tmp[2*i+1])
  			}
  		}
  	}
  	if n != 0 {
  		return FormatError("DQT has wrong length")
  	}
  	return nil
  }
  
  // Specified in section B.2.4.4.
  func (d *decoder) processDRI(n int) error {
  	if n != 2 {
  		return FormatError("DRI has wrong length")
  	}
  	if err := d.readFull(d.tmp[:2]); err != nil {
  		return err
  	}
  	d.ri = int(d.tmp[0])<<8 + int(d.tmp[1])
  	return nil
  }
  
  func (d *decoder) processApp0Marker(n int) error {
  	if n < 5 {
  		return d.ignore(n)
  	}
  	if err := d.readFull(d.tmp[:5]); err != nil {
  		return err
  	}
  	n -= 5
  
  	d.jfif = d.tmp[0] == 'J' && d.tmp[1] == 'F' && d.tmp[2] == 'I' && d.tmp[3] == 'F' && d.tmp[4] == '\x00'
  
  	if n > 0 {
  		return d.ignore(n)
  	}
  	return nil
  }
  
  func (d *decoder) processApp14Marker(n int) error {
  	if n < 12 {
  		return d.ignore(n)
  	}
  	if err := d.readFull(d.tmp[:12]); err != nil {
  		return err
  	}
  	n -= 12
  
  	if d.tmp[0] == 'A' && d.tmp[1] == 'd' && d.tmp[2] == 'o' && d.tmp[3] == 'b' && d.tmp[4] == 'e' {
  		d.adobeTransformValid = true
  		d.adobeTransform = d.tmp[11]
  	}
  
  	if n > 0 {
  		return d.ignore(n)
  	}
  	return nil
  }
  
  // decode reads a JPEG image from r and returns it as an image.Image.
  func (d *decoder) decode(r io.Reader, configOnly bool) (image.Image, error) {
  	d.r = r
  
  	// Check for the Start Of Image marker.
  	if err := d.readFull(d.tmp[:2]); err != nil {
  		return nil, err
  	}
  	if d.tmp[0] != 0xff || d.tmp[1] != soiMarker {
  		return nil, FormatError("missing SOI marker")
  	}
  
  	// Process the remaining segments until the End Of Image marker.
  	for {
  		err := d.readFull(d.tmp[:2])
  		if err != nil {
  			return nil, err
  		}
  		for d.tmp[0] != 0xff {
  			// Strictly speaking, this is a format error. However, libjpeg is
  			// liberal in what it accepts. As of version 9, next_marker in
  			// jdmarker.c treats this as a warning (JWRN_EXTRANEOUS_DATA) and
  			// continues to decode the stream. Even before next_marker sees
  			// extraneous data, jpeg_fill_bit_buffer in jdhuff.c reads as many
  			// bytes as it can, possibly past the end of a scan's data. It
  			// effectively puts back any markers that it overscanned (e.g. an
  			// "\xff\xd9" EOI marker), but it does not put back non-marker data,
  			// and thus it can silently ignore a small number of extraneous
  			// non-marker bytes before next_marker has a chance to see them (and
  			// print a warning).
  			//
  			// We are therefore also liberal in what we accept. Extraneous data
  			// is silently ignored.
  			//
  			// This is similar to, but not exactly the same as, the restart
  			// mechanism within a scan (the RST[0-7] markers).
  			//
  			// Note that extraneous 0xff bytes in e.g. SOS data are escaped as
  			// "\xff\x00", and so are detected a little further down below.
  			d.tmp[0] = d.tmp[1]
  			d.tmp[1], err = d.readByte()
  			if err != nil {
  				return nil, err
  			}
  		}
  		marker := d.tmp[1]
  		if marker == 0 {
  			// Treat "\xff\x00" as extraneous data.
  			continue
  		}
  		for marker == 0xff {
  			// Section B.1.1.2 says, "Any marker may optionally be preceded by any
  			// number of fill bytes, which are bytes assigned code X'FF'".
  			marker, err = d.readByte()
  			if err != nil {
  				return nil, err
  			}
  		}
  		if marker == eoiMarker { // End Of Image.
  			break
  		}
  		if rst0Marker <= marker && marker <= rst7Marker {
  			// Figures B.2 and B.16 of the specification suggest that restart markers should
  			// only occur between Entropy Coded Segments and not after the final ECS.
  			// However, some encoders may generate incorrect JPEGs with a final restart
  			// marker. That restart marker will be seen here instead of inside the processSOS
  			// method, and is ignored as a harmless error. Restart markers have no extra data,
  			// so we check for this before we read the 16-bit length of the segment.
  			continue
  		}
  
  		// Read the 16-bit length of the segment. The value includes the 2 bytes for the
  		// length itself, so we subtract 2 to get the number of remaining bytes.
  		if err = d.readFull(d.tmp[:2]); err != nil {
  			return nil, err
  		}
  		n := int(d.tmp[0])<<8 + int(d.tmp[1]) - 2
  		if n < 0 {
  			return nil, FormatError("short segment length")
  		}
  
  		switch marker {
  		case sof0Marker, sof1Marker, sof2Marker:
  			d.progressive = marker == sof2Marker
  			err = d.processSOF(n)
  			if configOnly && d.jfif {
  				return nil, err
  			}
  		case dhtMarker:
  			if configOnly {
  				err = d.ignore(n)
  			} else {
  				err = d.processDHT(n)
  			}
  		case dqtMarker:
  			if configOnly {
  				err = d.ignore(n)
  			} else {
  				err = d.processDQT(n)
  			}
  		case sosMarker:
  			if configOnly {
  				return nil, nil
  			}
  			err = d.processSOS(n)
  		case driMarker:
  			if configOnly {
  				err = d.ignore(n)
  			} else {
  				err = d.processDRI(n)
  			}
  		case app0Marker:
  			err = d.processApp0Marker(n)
  		case app14Marker:
  			err = d.processApp14Marker(n)
  		default:
  			if app0Marker <= marker && marker <= app15Marker || marker == comMarker {
  				err = d.ignore(n)
  			} else if marker < 0xc0 { // See Table B.1 "Marker code assignments".
  				err = FormatError("unknown marker")
  			} else {
  				err = UnsupportedError("unknown marker")
  			}
  		}
  		if err != nil {
  			return nil, err
  		}
  	}
  
  	if d.progressive {
  		if err := d.reconstructProgressiveImage(); err != nil {
  			return nil, err
  		}
  	}
  	if d.img1 != nil {
  		return d.img1, nil
  	}
  	if d.img3 != nil {
  		if d.blackPix != nil {
  			return d.applyBlack()
  		} else if d.isRGB() {
  			return d.convertToRGB()
  		}
  		return d.img3, nil
  	}
  	return nil, FormatError("missing SOS marker")
  }
  
  // applyBlack combines d.img3 and d.blackPix into a CMYK image. The formula
  // used depends on whether the JPEG image is stored as CMYK or YCbCrK,
  // indicated by the APP14 (Adobe) metadata.
  //
  // Adobe CMYK JPEG images are inverted, where 255 means no ink instead of full
  // ink, so we apply "v = 255 - v" at various points. Note that a double
  // inversion is a no-op, so inversions might be implicit in the code below.
  func (d *decoder) applyBlack() (image.Image, error) {
  	if !d.adobeTransformValid {
  		return nil, UnsupportedError("unknown color model: 4-component JPEG doesn't have Adobe APP14 metadata")
  	}
  
  	// If the 4-component JPEG image isn't explicitly marked as "Unknown (RGB
  	// or CMYK)" as per
  	// http://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/JPEG.html#Adobe
  	// we assume that it is YCbCrK. This matches libjpeg's jdapimin.c.
  	if d.adobeTransform != adobeTransformUnknown {
  		// Convert the YCbCr part of the YCbCrK to RGB, invert the RGB to get
  		// CMY, and patch in the original K. The RGB to CMY inversion cancels
  		// out the 'Adobe inversion' described in the applyBlack doc comment
  		// above, so in practice, only the fourth channel (black) is inverted.
  		bounds := d.img3.Bounds()
  		img := image.NewRGBA(bounds)
  		imageutil.DrawYCbCr(img, bounds, d.img3, bounds.Min)
  		for iBase, y := 0, bounds.Min.Y; y < bounds.Max.Y; iBase, y = iBase+img.Stride, y+1 {
  			for i, x := iBase+3, bounds.Min.X; x < bounds.Max.X; i, x = i+4, x+1 {
  				img.Pix[i] = 255 - d.blackPix[(y-bounds.Min.Y)*d.blackStride+(x-bounds.Min.X)]
  			}
  		}
  		return &image.CMYK{
  			Pix:    img.Pix,
  			Stride: img.Stride,
  			Rect:   img.Rect,
  		}, nil
  	}
  
  	// The first three channels (cyan, magenta, yellow) of the CMYK
  	// were decoded into d.img3, but each channel was decoded into a separate
  	// []byte slice, and some channels may be subsampled. We interleave the
  	// separate channels into an image.CMYK's single []byte slice containing 4
  	// contiguous bytes per pixel.
  	bounds := d.img3.Bounds()
  	img := image.NewCMYK(bounds)
  
  	translations := [4]struct {
  		src    []byte
  		stride int
  	}{
  		{d.img3.Y, d.img3.YStride},
  		{d.img3.Cb, d.img3.CStride},
  		{d.img3.Cr, d.img3.CStride},
  		{d.blackPix, d.blackStride},
  	}
  	for t, translation := range translations {
  		subsample := d.comp[t].h != d.comp[0].h || d.comp[t].v != d.comp[0].v
  		for iBase, y := 0, bounds.Min.Y; y < bounds.Max.Y; iBase, y = iBase+img.Stride, y+1 {
  			sy := y - bounds.Min.Y
  			if subsample {
  				sy /= 2
  			}
  			for i, x := iBase+t, bounds.Min.X; x < bounds.Max.X; i, x = i+4, x+1 {
  				sx := x - bounds.Min.X
  				if subsample {
  					sx /= 2
  				}
  				img.Pix[i] = 255 - translation.src[sy*translation.stride+sx]
  			}
  		}
  	}
  	return img, nil
  }
  
  func (d *decoder) isRGB() bool {
  	if d.jfif {
  		return false
  	}
  	if d.adobeTransformValid && d.adobeTransform == adobeTransformUnknown {
  		// http://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/JPEG.html#Adobe
  		// says that 0 means Unknown (and in practice RGB) and 1 means YCbCr.
  		return true
  	}
  	return d.comp[0].c == 'R' && d.comp[1].c == 'G' && d.comp[2].c == 'B'
  }
  
  func (d *decoder) convertToRGB() (image.Image, error) {
  	cScale := d.comp[0].h / d.comp[1].h
  	bounds := d.img3.Bounds()
  	img := image.NewRGBA(bounds)
  	for y := bounds.Min.Y; y < bounds.Max.Y; y++ {
  		po := img.PixOffset(bounds.Min.X, y)
  		yo := d.img3.YOffset(bounds.Min.X, y)
  		co := d.img3.COffset(bounds.Min.X, y)
  		for i, iMax := 0, bounds.Max.X-bounds.Min.X; i < iMax; i++ {
  			img.Pix[po+4*i+0] = d.img3.Y[yo+i]
  			img.Pix[po+4*i+1] = d.img3.Cb[co+i/cScale]
  			img.Pix[po+4*i+2] = d.img3.Cr[co+i/cScale]
  			img.Pix[po+4*i+3] = 255
  		}
  	}
  	return img, nil
  }
  
  // Decode reads a JPEG image from r and returns it as an image.Image.
  func Decode(r io.Reader) (image.Image, error) {
  	var d decoder
  	return d.decode(r, false)
  }
  
  // DecodeConfig returns the color model and dimensions of a JPEG image without
  // decoding the entire image.
  func DecodeConfig(r io.Reader) (image.Config, error) {
  	var d decoder
  	if _, err := d.decode(r, true); err != nil {
  		return image.Config{}, err
  	}
  	switch d.nComp {
  	case 1:
  		return image.Config{
  			ColorModel: color.GrayModel,
  			Width:      d.width,
  			Height:     d.height,
  		}, nil
  	case 3:
  		cm := color.YCbCrModel
  		if d.isRGB() {
  			cm = color.RGBAModel
  		}
  		return image.Config{
  			ColorModel: cm,
  			Width:      d.width,
  			Height:     d.height,
  		}, nil
  	case 4:
  		return image.Config{
  			ColorModel: color.CMYKModel,
  			Width:      d.width,
  			Height:     d.height,
  		}, nil
  	}
  	return image.Config{}, FormatError("missing SOF marker")
  }
  
  func init() {
  	image.RegisterFormat("jpeg", "\xff\xd8", Decode, DecodeConfig)
  }
  

View as plain text