image/jpeg: decode progressive JPEGs.

To be clear, this supports decoding the bytes on the wire into an
in-memory image. There is no API change: jpeg.Decode will still not
return until the entire image is decoded.

The code is obviously more complicated, and costs around 10% in
performance on baseline JPEGs. The processSOS code could be cleaned up a
bit, and maybe some of that loss can be reclaimed, but I'll leave that
for follow-up CLs, to keep the diff for this one as small as possible.

Before:
BenchmarkDecode	    1000	   2855637 ns/op	  21.64 MB/s
After:
BenchmarkDecodeBaseline	     500	   3178960 ns/op	  19.44 MB/s
BenchmarkDecodeProgressive	     500	   4082640 ns/op	  15.14 MB/s

Fixes #3976.

The test data was generated by:
# Create intermediate files; cjpeg on Ubuntu 10.04 can't read PNG.
convert video-001.png video-001.bmp
convert video-005.gray.png video-005.gray.pgm
# Create new test files.
cjpeg -quality 100 -sample 1x1,1x1,1x1 -progressive video-001.bmp > video-001.progressive.jpeg
cjpeg -quality 50 -sample 2x2,1x1,1x1 video-001.bmp > video-001.q50.420.jpeg
cjpeg -quality 50 -sample 2x1,1x1,1x1 video-001.bmp > video-001.q50.422.jpeg
cjpeg -quality 50 -sample 1x1,1x1,1x1 video-001.bmp > video-001.q50.444.jpeg
cjpeg -quality 50 -sample 2x2,1x1,1x1 -progressive video-001.bmp > video-001.q50.420.progressive.jpeg
cjpeg -quality 50 -sample 2x1,1x1,1x1 -progressive video-001.bmp > video-001.q50.422.progressive.jpeg
cjpeg -quality 50 -sample 1x1,1x1,1x1 -progressive video-001.bmp > video-001.q50.444.progressive.jpeg
cjpeg -quality 50 video-005.gray.pgm > video-005.gray.q50.jpeg
cjpeg -quality 50 -progressive video-005.gray.pgm > video-005.gray.q50.progressive.jpeg
# Delete intermediate files.
rm video-001.bmp video-005.gray.pgm

R=r
CC=golang-dev
https://golang.org/cl/6684046

src/pkg/image/decode_test.go

@@ -31,6 +31,7 @@ var imageTests = []imageTest{
 	{"testdata/video-001.png", "testdata/video-001.5bpp.gif", 128 << 8},
 	// JPEG is a lossy format and hence needs a non-zero tolerance.
 	{"testdata/video-001.png", "testdata/video-001.jpeg", 8 << 8},
+	{"testdata/video-001.png", "testdata/video-001.progressive.jpeg", 8 << 8},
 	// Grayscale images.
 	{"testdata/video-005.gray.png", "testdata/video-005.gray.jpeg", 8 << 8},
 	{"testdata/video-005.gray.png", "testdata/video-005.gray.png", 0},

src/pkg/image/jpeg/huffman.go

@@ -15,9 +15,9 @@ const maxNumValues = 256
 // Bit stream for the Huffman decoder.
 // The n least significant bits of a form the unread bits, to be read in MSB to LSB order.
 type bits struct {
-	a int // accumulator.
-	n int // the number of unread bits in a.
-	m int // mask. m==1<<(n-1) when n>0, with m==0 when n==0.
+	a uint32 // accumulator.
+	m uint32 // mask. m==1<<(n-1) when n>0, with m==0 when n==0.
+	n int    // the number of unread bits in a.
 }
 
 // Huffman table decoder, specified in section C.
@@ -39,7 +39,7 @@ func (d *decoder) ensureNBits(n int) error {
 		if err != nil {
 			return err
 		}
-		d.b.a = d.b.a<<8 | int(c)
+		d.b.a = d.b.a<<8 | uint32(c)
 		d.b.n += 8
 		if d.b.m == 0 {
 			d.b.m = 1 << 7
@@ -69,7 +69,7 @@ func (d *decoder) receiveExtend(t uint8) (int, error) {
 	d.b.n -= int(t)
 	d.b.m >>= t
 	s := 1 << t
-	x := (d.b.a >> uint8(d.b.n)) & (s - 1)
+	x := int(d.b.a>>uint8(d.b.n)) & (s - 1)
 	if x < s>>1 {
 		x += ((-1) << t) + 1
 	}
@@ -92,8 +92,7 @@ func (d *decoder) processDHT(n int) error {
 			return FormatError("bad Tc value")
 		}
 		th := d.tmp[0] & 0x0f
-		const isBaseline = true // Progressive mode is not yet supported.
-		if th > maxTh || isBaseline && th > 1 {
+		if th > maxTh || !d.progressive && th > 1 {
 			return FormatError("bad Th value")
 		}
 		h := &d.huff[tc][th]
@@ -185,3 +184,28 @@ func (d *decoder) decodeHuffman(h *huffman) (uint8, error) {
 	}
 	return 0, FormatError("bad Huffman code")
 }
+
+func (d *decoder) decodeBit() (bool, error) {
+	if d.b.n == 0 {
+		err := d.ensureNBits(1)
+		if err != nil {
+			return false, err
+		}
+	}
+	ret := d.b.a&d.b.m != 0
+	d.b.n--
+	d.b.m >>= 1
+	return ret, nil
+}
+
+func (d *decoder) decodeBits(n int) (uint32, error) {
+	err := d.ensureNBits(n)
+	if err != nil {
+		return 0, err
+	}
+	ret := d.b.a >> uint(d.b.n-n)
+	ret &= (1 << uint(n)) - 1
+	d.b.n -= n
+	d.b.m >>= uint(n)
+	return ret, nil
+}

src/pkg/image/jpeg/reader_test.go

+// Copyright 2012 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
+
+import (
+	"bytes"
+	"fmt"
+	"image"
+	"io/ioutil"
+	"os"
+	"testing"
+)
+
+// TestDecodeProgressive tests that decoding the baseline and progressive
+// versions of the same image result in exactly the same pixel data, in YCbCr
+// space for color images, and Y space for grayscale images.
+func TestDecodeProgressive(t *testing.T) {
+	testCases := []string{
+		"../testdata/video-001",
+		"../testdata/video-001.q50.420",
+		"../testdata/video-001.q50.422",
+		"../testdata/video-001.q50.444",
+		"../testdata/video-005.gray.q50",
+	}
+	for _, tc := range testCases {
+		m0, err := decodeFile(tc + ".jpeg")
+		if err != nil {
+			t.Errorf("%s: %v", tc+".jpeg", err)
+			continue
+		}
+		m1, err := decodeFile(tc + ".progressive.jpeg")
+		if err != nil {
+			t.Errorf("%s: %v", tc+".progressive.jpeg", err)
+			continue
+		}
+		if m0.Bounds() != m1.Bounds() {
+			t.Errorf("%s: bounds differ: %v and %v", tc, m0.Bounds(), m1.Bounds())
+			continue
+		}
+		switch m0 := m0.(type) {
+		case *image.YCbCr:
+			m1 := m1.(*image.YCbCr)
+			if err := check(m0.Bounds(), m0.Y, m1.Y, m0.YStride, m1.YStride); err != nil {
+				t.Errorf("%s (Y): %v", tc, err)
+				continue
+			}
+			if err := check(m0.Bounds(), m0.Cb, m1.Cb, m0.CStride, m1.CStride); err != nil {
+				t.Errorf("%s (Cb): %v", tc, err)
+				continue
+			}
+			if err := check(m0.Bounds(), m0.Cr, m1.Cr, m0.CStride, m1.CStride); err != nil {
+				t.Errorf("%s (Cr): %v", tc, err)
+				continue
+			}
+		case *image.Gray:
+			m1 := m1.(*image.Gray)
+			if err := check(m0.Bounds(), m0.Pix, m1.Pix, m0.Stride, m1.Stride); err != nil {
+				t.Errorf("%s: %v", tc, err)
+				continue
+			}
+		default:
+			t.Errorf("%s: unexpected image type %T", tc, m0)
+			continue
+		}
+	}
+}
+
+func decodeFile(filename string) (image.Image, error) {
+	f, err := os.Open(filename)
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+	return Decode(f)
+
+}
+
+// check checks that the two pix data are equal, within the given bounds.
+func check(bounds image.Rectangle, pix0, pix1 []byte, stride0, stride1 int) error {
+	if len(pix0) != len(pix1) {
+		return fmt.Errorf("len(pix) %d and %d differ", len(pix0), len(pix1))
+	}
+	if stride0 != stride1 {
+		return fmt.Errorf("strides %d and %d differ", stride0, stride1)
+	}
+	if stride0%8 != 0 {
+		return fmt.Errorf("stride %d is not a multiple of 8", stride0)
+	}
+	// Compare the two pix data, one 8x8 block at a time.
+	for y := 0; y < len(pix0)/stride0; y += 8 {
+		for x := 0; x < stride0; x += 8 {
+			if x >= bounds.Max.X || y >= bounds.Max.Y {
+				// We don't care if the two pix data differ if the 8x8 block is
+				// entirely outside of the image's bounds. For example, this can
+				// occur with a 4:2:0 chroma subsampling and a 1x1 image. Baseline
+				// decoding works on the one 16x16 MCU as a whole; progressive
+				// decoding's first pass works on that 16x16 MCU as a whole but
+				// refinement passes only process one 8x8 block within the MCU.
+				continue
+			}
+
+			for j := 0; j < 8; j++ {
+				for i := 0; i < 8; i++ {
+					index := (y+j)*stride0 + (x + i)
+					if pix0[index] != pix1[index] {
+						return fmt.Errorf("blocks at (%d, %d) differ:\n%sand\n%s", x, y,
+							pixString(pix0, stride0, x, y),
+							pixString(pix1, stride1, x, y),
+						)
+					}
+				}
+			}
+		}
+	}
+	return nil
+}
+
+func pixString(pix []byte, stride, x, y int) string {
+	s := bytes.NewBuffer(nil)
+	for j := 0; j < 8; j++ {
+		fmt.Fprintf(s, "\t")
+		for i := 0; i < 8; i++ {
+			fmt.Fprintf(s, "%02x ", pix[(y+j)*stride+(x+i)])
+		}
+		fmt.Fprintf(s, "\n")
+	}
+	return s.String()
+}
+
+func benchmarkDecode(b *testing.B, filename string) {
+	b.StopTimer()
+	data, err := ioutil.ReadFile(filename)
+	if err != nil {
+		b.Fatal(err)
+	}
+	cfg, err := DecodeConfig(bytes.NewReader(data))
+	if err != nil {
+		b.Fatal(err)
+	}
+	b.SetBytes(int64(cfg.Width * cfg.Height * 4))
+	b.StartTimer()
+	for i := 0; i < b.N; i++ {
+		Decode(bytes.NewReader(data))
+	}
+}
+
+func BenchmarkDecodeBaseline(b *testing.B) {
+	benchmarkDecode(b, "../testdata/video-001.jpeg")
+}
+
+func BenchmarkDecodeProgressive(b *testing.B) {
+	benchmarkDecode(b, "../testdata/video-001.progressive.jpeg")
+}

src/pkg/image/jpeg/scan.go

+// Copyright 2012 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
+
+// refine decodes a successive approximation refinement block, as specified in
+// section G.1.2.
+func (d *decoder) refine(b *block, h *huffman, zigStart, zigEnd, delta int) error {
+	// Refining a DC component is trivial.
+	if zigStart == 0 {
+		if zigEnd != 0 {
+			panic("unreachable")
+		}
+		bit, err := d.decodeBit()
+		if err != nil {
+			return err
+		}
+		if bit {
+			b[0] |= delta
+		}
+		return nil
+	}
+
+	// Refining AC components is more complicated; see sections G.1.2.2 and G.1.2.3.
+	zig := zigStart
+	if d.eobRun == 0 {
+	loop:
+		for ; zig <= zigEnd; zig++ {
+			z := 0
+			value, err := d.decodeHuffman(h)
+			if err != nil {
+				return err
+			}
+			val0 := value >> 4
+			val1 := value & 0x0f
+
+			switch val1 {
+			case 0:
+				if val0 != 0x0f {
+					d.eobRun = uint16(1 << val0)
+					if val0 != 0 {
+						bits, err := d.decodeBits(int(val0))
+						if err != nil {
+							return err
+						}
+						d.eobRun |= uint16(bits)
+					}
+					break loop
+				}
+			case 1:
+				z = delta
+				bit, err := d.decodeBit()
+				if err != nil {
+					return err
+				}
+				if !bit {
+					z = -z
+				}
+			default:
+				return FormatError("unexpected Huffman code")
+			}
+
+			zig, err = d.refineNonZeroes(b, zig, zigEnd, int(val0), delta)
+			if err != nil {
+				return err
+			}
+			if zig > zigEnd {
+				return FormatError("too many coefficients")
+			}
+			if z != 0 {
+				b[unzig[zig]] = z
+			}
+		}
+	}
+	if d.eobRun > 0 {
+		d.eobRun--
+		if _, err := d.refineNonZeroes(b, zig, zigEnd, -1, delta); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// refineNonZeroes refines non-zero entries of b in zig-zag order. If nz >= 0,
+// the first nz zero entries are skipped over.
+func (d *decoder) refineNonZeroes(b *block, zig, zigEnd, nz, delta int) (int, error) {
+	for ; zig <= zigEnd; zig++ {
+		u := unzig[zig]
+		if b[u] == 0 {
+			if nz == 0 {
+				break
+			}
+			nz--
+			continue
+		}
+		bit, err := d.decodeBit()
+		if err != nil {
+			return 0, err
+		}
+		if !bit {
+			continue
+		}
+		if b[u] >= 0 {
+			b[u] += delta
+		} else {
+			b[u] -= delta
+		}
+	}
+	return zig, nil
+}

src/pkg/image/jpeg/writer_test.go

@@ -171,23 +171,6 @@ func TestWriter(t *testing.T) {
 	}
 }
 
-func BenchmarkDecode(b *testing.B) {
-	b.StopTimer()
-	data, err := ioutil.ReadFile("../testdata/video-001.jpeg")
-	if err != nil {
-		b.Fatal(err)
-	}
-	cfg, err := DecodeConfig(bytes.NewReader(data))
-	if err != nil {
-		b.Fatal(err)
-	}
-	b.SetBytes(int64(cfg.Width * cfg.Height * 4))
-	b.StartTimer()
-	for i := 0; i < b.N; i++ {
-		Decode(bytes.NewReader(data))
-	}
-}
-
 func BenchmarkEncode(b *testing.B) {
 	b.StopTimer()
 	img := image.NewRGBA(image.Rect(0, 0, 640, 480))