The Independent JPEG Group's JPEG software v6b with arithmetic coding support
diff --git a/README.arithmetic b/README.arithmetic
new file mode 100644
index 0000000..0dd3b9a
--- /dev/null
+++ b/README.arithmetic
@@ -0,0 +1,215 @@
+JPEG arithmetic encoding and decoding portable software implementation
+======================================================================
+
+Release of 28-Mar-98 by Guido Vollbeding <guido@jpegclub.org>
+=============================================================
+
+Primary URLs:
+
+ http://sylvana.net/jpeg-ari/
+ (directory containing the actual archive files:)
+
+ http://sylvana.net/jpeg-ari/jpeg-ari-28mar98.tar.gz
+
+ http://sylvana.net/jpeg-ari/jpeg-ari.zip
+
+
+DISCLAIMER
+==========
+
+This package is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+
+It is possible that certain products which can be built using this
+software modules might form inventions protected by patent rights in
+some countries (e.g. by patents about arithmetic coding algorithms
+owned by IBM and AT&T in the USA). Provision of this software by the
+author does NOT include any licenses for any patents.
+In those countries where a patent license is required for certain
+applications of this software modules, you will have to obtain such
+a license yourself.
+
+See Annex L in the JPEG spec for further information
+and a list of relevant patents.
+
+
+What is it?
+===========
+
+This is my implementation of the arithmetic encoding and decoding
+back-end for JPEG as specified in the
+
+ ISO/IEC International Standard 10918-1 and CCITT Recommendation
+ ITU-T T.81, "Information Technology - Digital Compression and
+ Coding of Continuous-tone Still Images, Part 1: Requirements
+ and Guidelines".
+
+Arithmetic coding is a state-of-the-art lossless entropy data
+compression method which offers better compression performance
+than the well-established Huffman entropy coding process.
+
+The JPEG standard specifies a particular arithmetic coding scheme
+to be used optionally as alternative to Huffman coding.
+
+
+Who needs it?
+=============
+
+This package might be of interest for people who are looking for
+enhanced state-of-the-art image compression technologies.
+
+It is intended to provide a reasonable tool for experimental,
+comparison and evaluation purposes.
+
+See the Disclaimer above for restricted conditions of usage.
+
+
+How does it work?
+=================
+
+This distribution is organized as add-on to the widespread
+Independent JPEG Group's JPEG software.
+
+Thus, once you managed to install the IJG software distribution
+successfully, there should be no additional problems (portability
+issues etc.) to incorporate this package into the library,
+and usage is straightforward.
+
+Transcode given JPEG files simply with a command like
+
+ jpegtran -arithmetic [-progressive] < orig.jpg > arit.jpg
+
+into an arithmetic coded version LOSSLESSLY! Since there are
+practically no applications in existence which can handle such
+files, you can only transform it back with the same tool
+
+ jpegtran [-optimize] [-progressive] < arit.jpg > orig2.jpg
+
+to verify correct operation.
+
+Thus, you can easily verify the enhanced compression performance
+of the arithmetic coding version compared to the Huffman (with
+fixed or custom tables) version.
+
+The claim to evaluate was that arithmetic coding gives an average
+5-10% compression improvement against Huffman.
+Early tests with this implementation support this claim, and you
+can perform tests with own material.
+
+Here are some actual results:
+
+% ./jpegtran -optimize < testorig.jpg > testopt.jpg
+% ./jpegtran -arithmetic < testorig.jpg > testarit.jpg
+% ./jpegtran < testarit.jpg > testorig2.jpg
+% ./jpegtran -arithmetic -progressive < testorig.jpg > testaritp.jpg
+% ./jpegtran < testaritp.jpg > testorig3.jpg
+% ./jpegtran -optimize < ../butterfly.jpg > ../buttopt.jpg
+% ./jpegtran -progressive < ../butterfly.jpg > ../buttprog.jpg
+% ./jpegtran -arithmetic < ../butterfly.jpg > ../buttarit.jpg
+% ./jpegtran < ../buttarit.jpg > ../butterfly2.jpg
+% ./jpegtran -arithmetic -progressive < ../butterfly.jpg > ../buttaritp.jpg
+% ./jpegtran < ../buttaritp.jpg > ../butterfly3.jpg
+% ls -l test*.jpg
+-rw-r--r-- 1 guivol 5153 Apr 13 18:51 testarit.jpg
+-rw-r--r-- 1 guivol 5186 Apr 13 18:51 testaritp.jpg
+-rw-r--r-- 1 guivol 5756 Apr 2 15:10 testimg.jpg
+-rw-r--r-- 1 guivol 5645 Apr 2 15:10 testimgp.jpg
+-rw-r--r-- 1 guivol 5463 Apr 13 18:51 testopt.jpg
+-rw-r--r-- 1 guivol 5770 Apr 2 15:10 testorig.jpg
+-rw-r--r-- 1 guivol 5770 Apr 13 18:51 testorig2.jpg
+-rw-r--r-- 1 guivol 5770 Apr 13 18:51 testorig3.jpg
+-rw-r--r-- 1 guivol 5655 Apr 2 15:10 testprog.jpg
+% ls -l ../butt*.jpg
+-rw-r--r-- 1 guivol 460091 Apr 13 18:52 ../buttarit.jpg
+-rw-r--r-- 1 guivol 453703 Apr 13 18:52 ../buttaritp.jpg
+-rw-r--r-- 1 guivol 527823 Nov 19 18:41 ../butterfly.jpg
+-rw-r--r-- 1 guivol 527823 Apr 13 18:52 ../butterfly2.jpg
+-rw-r--r-- 1 guivol 527823 Apr 13 18:52 ../butterfly3.jpg
+-rw-r--r-- 1 guivol 511834 Apr 13 18:52 ../buttopt.jpg
+-rw-r--r-- 1 guivol 492237 Apr 13 18:52 ../buttprog.jpg
+%
+
+Note that arithmetic coding requires only a single processing
+pass due to its fully-adaptive nature, and compared to one-pass
+(fixed tables) Huffman the arithmetic coded version consistently
+achieves 10% compression improvement.
+Compared with two-pass (custom tables) Huffman the improvement
+is 5-10%.
+
+Note that I wasn't able yet to cross-check interoperability of
+the produced files with other implementations.
+Thus, I can't be sure that the files are compliant to the spec,
+but I hope so and the tests support it.
+The encoding and decoding processes should be correct anyway,
+however, in the sense that they are complementary to each other
+and thus retain data integrity.
+
+I would appreciate any indications for compliance or interoperability
+with other implementations from somebody.
+Please let me know if you are able to cross-check something.
+
+
+Installation
+============
+
+The installation is a 2-stage procedure:
+
+1. Preparing the IJG package for potential incorporation
+ of the arithmetic coding feature.
+
+2. Incorporation of the actual arithmetic coding modules
+ and enabling the feature for usage.
+
+The reason for this 2-stage process is the hope to make
+step 1 obsolete in future IJG releases.
+The actual implementation should remain separate IMHO due
+to the different usage conditions.
+
+Step 1:
+
+1.1. Copy all files from the subdirectory 'patchv6b' into
+ the IJG software's v6b source directory.
+ This includes minor patches to some files and 3 extra
+ files which hold place for the actual implementation.
+
+1.2. Update your Makefile/Projectfile for the inclusion of
+ the 3 extra files. This will be done automatically
+ if you use a configure-generated makefile and type
+ './configure' (reconfigure).
+
+1.3. Recompile ('make').
+
+See the file 'PATCHES' in 'patchv6b' for details.
+
+Step 2:
+
+2.1. Replace the 3 placeholder files by the actual implementation
+ modules.
+
+2.2. Enable application support of the new features by #defining
+ C_ARITH_CODING_SUPPORTED and D_ARITH_CODING_SUPPORTED
+ in 'jmorecfg.h'.
+
+2.3. Recompile ('make').
+
+Note that I suggest to add 3 placeholder files to the IJG
+distribution. This would remove the need for system-dependent
+changes (Makefiles) and thus considerably simplify the actual
+installation for systems without a configure-generated makefile.
+
+
+References
+==========
+
+- The Independent JPEG Group's software
+
+- JBIG-KIT lossless image compression library by Markus Kuhn
+
+- William B. Pennebaker, Joan L. Mitchell:
+ "JPEG Still Image Data Compression Standard",
+ Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
+
+- jpeg-faq (http://www.faqs.org/faqs/jpeg-faq/)
+
+- compression-faq (http://www.faqs.org/faqs/compression-faq/)
diff --git a/jaricom.c b/jaricom.c
new file mode 100644
index 0000000..e4292b0
--- /dev/null
+++ b/jaricom.c
@@ -0,0 +1,149 @@
+/*
+ * jaricom.c
+ *
+ * Copyright (C) 1997, Guido Vollbeding <guivol@esc.de>.
+ * This file is NOT part of the Independent JPEG Group's software
+ * for legal reasons.
+ * See the accompanying README file for conditions of distribution and use.
+ *
+ * This file contains probability estimation tables for common use in
+ * arithmetic entropy encoding and decoding routines.
+ *
+ * This data represents Table D.2 in the JPEG spec (ISO/IEC IS 10918-1
+ * and CCITT Recommendation ITU-T T.81) and Table 24 in the JBIG spec
+ * (ISO/IEC IS 11544 and CCITT Recommendation ITU-T T.82).
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+/* The following #define specifies the packing of the four components
+ * into the compact INT32 representation.
+ * Note that this formula must match the actual arithmetic encoder
+ * and decoder implementation. The implementation has to be changed
+ * if this formula is changed.
+ * The current organisation is leaned on Markus Kuhn's JBIG
+ * implementation (jbig_tab.c).
+ */
+
+#define V(a,b,c,d) (((INT32)a << 16) | ((INT32)c << 8) | ((INT32)d << 7) | b)
+
+const INT32 jaritab[113] = {
+/*
+ * Index, Qe_Value, Next_Index_LPS, Next_Index_MPS, Switch_MPS
+ */
+/* 0 */ V( 0x5a1d, 1, 1, 1 ),
+/* 1 */ V( 0x2586, 14, 2, 0 ),
+/* 2 */ V( 0x1114, 16, 3, 0 ),
+/* 3 */ V( 0x080b, 18, 4, 0 ),
+/* 4 */ V( 0x03d8, 20, 5, 0 ),
+/* 5 */ V( 0x01da, 23, 6, 0 ),
+/* 6 */ V( 0x00e5, 25, 7, 0 ),
+/* 7 */ V( 0x006f, 28, 8, 0 ),
+/* 8 */ V( 0x0036, 30, 9, 0 ),
+/* 9 */ V( 0x001a, 33, 10, 0 ),
+/* 10 */ V( 0x000d, 35, 11, 0 ),
+/* 11 */ V( 0x0006, 9, 12, 0 ),
+/* 12 */ V( 0x0003, 10, 13, 0 ),
+/* 13 */ V( 0x0001, 12, 13, 0 ),
+/* 14 */ V( 0x5a7f, 15, 15, 1 ),
+/* 15 */ V( 0x3f25, 36, 16, 0 ),
+/* 16 */ V( 0x2cf2, 38, 17, 0 ),
+/* 17 */ V( 0x207c, 39, 18, 0 ),
+/* 18 */ V( 0x17b9, 40, 19, 0 ),
+/* 19 */ V( 0x1182, 42, 20, 0 ),
+/* 20 */ V( 0x0cef, 43, 21, 0 ),
+/* 21 */ V( 0x09a1, 45, 22, 0 ),
+/* 22 */ V( 0x072f, 46, 23, 0 ),
+/* 23 */ V( 0x055c, 48, 24, 0 ),
+/* 24 */ V( 0x0406, 49, 25, 0 ),
+/* 25 */ V( 0x0303, 51, 26, 0 ),
+/* 26 */ V( 0x0240, 52, 27, 0 ),
+/* 27 */ V( 0x01b1, 54, 28, 0 ),
+/* 28 */ V( 0x0144, 56, 29, 0 ),
+/* 29 */ V( 0x00f5, 57, 30, 0 ),
+/* 30 */ V( 0x00b7, 59, 31, 0 ),
+/* 31 */ V( 0x008a, 60, 32, 0 ),
+/* 32 */ V( 0x0068, 62, 33, 0 ),
+/* 33 */ V( 0x004e, 63, 34, 0 ),
+/* 34 */ V( 0x003b, 32, 35, 0 ),
+/* 35 */ V( 0x002c, 33, 9, 0 ),
+/* 36 */ V( 0x5ae1, 37, 37, 1 ),
+/* 37 */ V( 0x484c, 64, 38, 0 ),
+/* 38 */ V( 0x3a0d, 65, 39, 0 ),
+/* 39 */ V( 0x2ef1, 67, 40, 0 ),
+/* 40 */ V( 0x261f, 68, 41, 0 ),
+/* 41 */ V( 0x1f33, 69, 42, 0 ),
+/* 42 */ V( 0x19a8, 70, 43, 0 ),
+/* 43 */ V( 0x1518, 72, 44, 0 ),
+/* 44 */ V( 0x1177, 73, 45, 0 ),
+/* 45 */ V( 0x0e74, 74, 46, 0 ),
+/* 46 */ V( 0x0bfb, 75, 47, 0 ),
+/* 47 */ V( 0x09f8, 77, 48, 0 ),
+/* 48 */ V( 0x0861, 78, 49, 0 ),
+/* 49 */ V( 0x0706, 79, 50, 0 ),
+/* 50 */ V( 0x05cd, 48, 51, 0 ),
+/* 51 */ V( 0x04de, 50, 52, 0 ),
+/* 52 */ V( 0x040f, 50, 53, 0 ),
+/* 53 */ V( 0x0363, 51, 54, 0 ),
+/* 54 */ V( 0x02d4, 52, 55, 0 ),
+/* 55 */ V( 0x025c, 53, 56, 0 ),
+/* 56 */ V( 0x01f8, 54, 57, 0 ),
+/* 57 */ V( 0x01a4, 55, 58, 0 ),
+/* 58 */ V( 0x0160, 56, 59, 0 ),
+/* 59 */ V( 0x0125, 57, 60, 0 ),
+/* 60 */ V( 0x00f6, 58, 61, 0 ),
+/* 61 */ V( 0x00cb, 59, 62, 0 ),
+/* 62 */ V( 0x00ab, 61, 63, 0 ),
+/* 63 */ V( 0x008f, 61, 32, 0 ),
+/* 64 */ V( 0x5b12, 65, 65, 1 ),
+/* 65 */ V( 0x4d04, 80, 66, 0 ),
+/* 66 */ V( 0x412c, 81, 67, 0 ),
+/* 67 */ V( 0x37d8, 82, 68, 0 ),
+/* 68 */ V( 0x2fe8, 83, 69, 0 ),
+/* 69 */ V( 0x293c, 84, 70, 0 ),
+/* 70 */ V( 0x2379, 86, 71, 0 ),
+/* 71 */ V( 0x1edf, 87, 72, 0 ),
+/* 72 */ V( 0x1aa9, 87, 73, 0 ),
+/* 73 */ V( 0x174e, 72, 74, 0 ),
+/* 74 */ V( 0x1424, 72, 75, 0 ),
+/* 75 */ V( 0x119c, 74, 76, 0 ),
+/* 76 */ V( 0x0f6b, 74, 77, 0 ),
+/* 77 */ V( 0x0d51, 75, 78, 0 ),
+/* 78 */ V( 0x0bb6, 77, 79, 0 ),
+/* 79 */ V( 0x0a40, 77, 48, 0 ),
+/* 80 */ V( 0x5832, 80, 81, 1 ),
+/* 81 */ V( 0x4d1c, 88, 82, 0 ),
+/* 82 */ V( 0x438e, 89, 83, 0 ),
+/* 83 */ V( 0x3bdd, 90, 84, 0 ),
+/* 84 */ V( 0x34ee, 91, 85, 0 ),
+/* 85 */ V( 0x2eae, 92, 86, 0 ),
+/* 86 */ V( 0x299a, 93, 87, 0 ),
+/* 87 */ V( 0x2516, 86, 71, 0 ),
+/* 88 */ V( 0x5570, 88, 89, 1 ),
+/* 89 */ V( 0x4ca9, 95, 90, 0 ),
+/* 90 */ V( 0x44d9, 96, 91, 0 ),
+/* 91 */ V( 0x3e22, 97, 92, 0 ),
+/* 92 */ V( 0x3824, 99, 93, 0 ),
+/* 93 */ V( 0x32b4, 99, 94, 0 ),
+/* 94 */ V( 0x2e17, 93, 86, 0 ),
+/* 95 */ V( 0x56a8, 95, 96, 1 ),
+/* 96 */ V( 0x4f46, 101, 97, 0 ),
+/* 97 */ V( 0x47e5, 102, 98, 0 ),
+/* 98 */ V( 0x41cf, 103, 99, 0 ),
+/* 99 */ V( 0x3c3d, 104, 100, 0 ),
+/* 100 */ V( 0x375e, 99, 93, 0 ),
+/* 101 */ V( 0x5231, 105, 102, 0 ),
+/* 102 */ V( 0x4c0f, 106, 103, 0 ),
+/* 103 */ V( 0x4639, 107, 104, 0 ),
+/* 104 */ V( 0x415e, 103, 99, 0 ),
+/* 105 */ V( 0x5627, 105, 106, 1 ),
+/* 106 */ V( 0x50e7, 108, 107, 0 ),
+/* 107 */ V( 0x4b85, 109, 103, 0 ),
+/* 108 */ V( 0x5597, 110, 109, 0 ),
+/* 109 */ V( 0x504f, 111, 107, 0 ),
+/* 110 */ V( 0x5a10, 110, 111, 1 ),
+/* 111 */ V( 0x5522, 112, 109, 0 ),
+/* 112 */ V( 0x59eb, 112, 111, 1 )
+};
diff --git a/jcarith.c b/jcarith.c
new file mode 100644
index 0000000..ae1b23c
--- /dev/null
+++ b/jcarith.c
@@ -0,0 +1,922 @@
+/*
+ * jcarith.c
+ *
+ * Copyright (C) 1997, Guido Vollbeding <guivol@esc.de>.
+ * This file is NOT part of the Independent JPEG Group's software
+ * for legal reasons.
+ * See the accompanying README file for conditions of distribution and use.
+ *
+ * This file contains portable arithmetic entropy encoding routines for JPEG
+ * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
+ *
+ * Both sequential and progressive modes are supported in this single module.
+ *
+ * Suspension is not currently supported in this module.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Expanded entropy encoder object for arithmetic encoding. */
+
+typedef struct {
+ struct jpeg_entropy_encoder pub; /* public fields */
+
+ INT32 c; /* C register, base of coding interval, layout as in sec. D.1.3 */
+ INT32 a; /* A register, normalized size of coding interval */
+ INT32 sc; /* counter for stacked 0xFF values which might overflow */
+ INT32 zc; /* counter for pending 0x00 output values which might *
+ * be discarded at the end ("Pacman" termination) */
+ int ct; /* bit shift counter, determines when next byte will be written */
+ int buffer; /* buffer for most recent output byte != 0xFF */
+
+ int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+ int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
+
+ unsigned int restarts_to_go; /* MCUs left in this restart interval */
+ int next_restart_num; /* next restart number to write (0-7) */
+
+ /* Pointers to statistics areas (these workspaces have image lifespan) */
+ unsigned char * dc_stats[NUM_ARITH_TBLS];
+ unsigned char * ac_stats[NUM_ARITH_TBLS];
+} arith_entropy_encoder;
+
+typedef arith_entropy_encoder * arith_entropy_ptr;
+
+/* The following two definitions specify the allocation chunk size
+ * for the statistics area.
+ * According to sections F.1.4.4.1.3 and F.1.4.4.2, we need at least
+ * 49 statistics bins for DC, and 245 statistics bins for AC coding.
+ * Note that we use one additional AC bin for codings with fixed
+ * probability (0.5), thus the minimum number for AC is 246.
+ *
+ * We use a compact representation with 1 byte per statistics bin,
+ * thus the numbers directly represent byte sizes.
+ * This 1 byte per statistics bin contains the meaning of the MPS
+ * (more probable symbol) in the highest bit (mask 0x80), and the
+ * index into the probability estimation state machine table
+ * in the lower bits (mask 0x7F).
+ */
+
+#define DC_STAT_BINS 64
+#define AC_STAT_BINS 256
+
+/* NOTE: Uncomment the following #define if you want to use the
+ * given formula for calculating the AC conditioning parameter Kx
+ * for spectral selection progressive coding in section G.1.3.2
+ * of the spec (Kx = Kmin + SRL (8 + Se - Kmin) 4).
+ * Although the spec and P&M authors claim that this "has proven
+ * to give good results for 8 bit precision samples", I'm not
+ * convinced yet that this is really beneficial.
+ * Early tests gave only very marginal compression enhancements
+ * (a few - around 5 or so - bytes even for very large files),
+ * which would turn out rather negative if we'd suppress the
+ * DAC (Define Arithmetic Conditioning) marker segments for
+ * the default parameters in the future.
+ * Note that currently the marker writing module emits 12-byte
+ * DAC segments for a full-component scan in a color image.
+ * This is not worth worrying about IMHO. However, since the
+ * spec defines the default values to be used if the tables
+ * are omitted (unlike Huffman tables, which are required
+ * anyway), one might optimize this behaviour in the future,
+ * and then it would be disadvantageous to use custom tables if
+ * they don't provide sufficient gain to exceed the DAC size.
+ *
+ * On the other hand, I'd consider it as a reasonable result
+ * that the conditioning has no significant influence on the
+ * compression performance. This means that the basic
+ * statistical model is already rather stable.
+ *
+ * Thus, at the moment, we use the default conditioning values
+ * anyway, and do not use the custom formula.
+ *
+#define CALCULATE_SPECTRAL_CONDITIONING
+ */
+
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
+ * We assume that int right shift is unsigned if INT32 right shift is,
+ * which should be safe.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS int ishift_temp;
+#define IRIGHT_SHIFT(x,shft) \
+ ((ishift_temp = (x)) < 0 ? \
+ (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+ (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x,shft) ((x) >> (shft))
+#endif
+
+
+LOCAL(void)
+emit_byte (int val, j_compress_ptr cinfo)
+/* Write next output byte; we do not support suspension in this module. */
+{
+ struct jpeg_destination_mgr * dest = cinfo->dest;
+
+ *dest->next_output_byte++ = (JOCTET) val;
+ if (--dest->free_in_buffer == 0)
+ if (! (*dest->empty_output_buffer) (cinfo))
+ ERREXIT(cinfo, JERR_CANT_SUSPEND);
+}
+
+
+/*
+ * Finish up at the end of an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+finish_pass (j_compress_ptr cinfo)
+{
+ arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
+ INT32 temp;
+
+ /* Section D.1.8: Termination of encoding */
+
+ /* Find the e->c in the coding interval with the largest
+ * number of trailing zero bits */
+ if ((temp = (e->a - 1 + e->c) & 0xFFFF0000L) < e->c)
+ e->c = temp + 0x8000L;
+ else
+ e->c = temp;
+ /* Send remaining bytes to output */
+ e->c <<= e->ct;
+ if (e->c & 0xF8000000L) {
+ /* One final overflow has to be handled */
+ if (e->buffer >= 0) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ emit_byte(e->buffer + 1, cinfo);
+ if (e->buffer + 1 == 0xFF)
+ emit_byte(0x00, cinfo);
+ }
+ e->zc += e->sc; /* carry-over converts stacked 0xFF bytes to 0x00 */
+ e->sc = 0;
+ } else {
+ if (e->buffer == 0)
+ ++e->zc;
+ else if (e->buffer >= 0) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ emit_byte(e->buffer, cinfo);
+ }
+ if (e->sc) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ do {
+ emit_byte(0xFF, cinfo);
+ emit_byte(0x00, cinfo);
+ } while (--e->sc);
+ }
+ }
+ /* Output final bytes only if they are not 0x00 */
+ if (e->c & 0x7FFF800L) {
+ if (e->zc) /* output final pending zero bytes */
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ emit_byte((e->c >> 19) & 0xFF, cinfo);
+ if (((e->c >> 19) & 0xFF) == 0xFF)
+ emit_byte(0x00, cinfo);
+ if (e->c & 0x7F800L) {
+ emit_byte((e->c >> 11) & 0xFF, cinfo);
+ if (((e->c >> 11) & 0xFF) == 0xFF)
+ emit_byte(0x00, cinfo);
+ }
+ }
+}
+
+
+/*
+ * The core arithmetic encoding routine (common in JPEG and JBIG).
+ * This needs to go as fast as possible.
+ * Machine-dependent optimization facilities
+ * are not utilized in this portable implementation.
+ * However, this code should be fairly efficient and
+ * may be a good base for further optimizations anyway.
+ *
+ * Parameter 'val' to be encoded may be 0 or 1 (binary decision).
+ *
+ * Note: I've added full "Pacman" termination support to the
+ * byte output routines, which is equivalent to the optional
+ * Discard_final_zeros procedure (Figure D.15) in the spec.
+ * Thus, we always produce the shortest possible output
+ * stream compliant to the spec (no trailing zero bytes,
+ * except for FF stuffing).
+ *
+ * I've also introduced a new scheme for accessing
+ * the probability estimation state machine table,
+ * derived from Markus Kuhn's JBIG implementation.
+ */
+
+LOCAL(void)
+arith_encode (j_compress_ptr cinfo, unsigned char *st, int val)
+{
+ extern const INT32 jaritab[];
+ register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
+ register unsigned char nl, nm;
+ register INT32 qe, temp;
+ register int sv;
+
+ /* Fetch values from our compact representation of Table D.2:
+ * Qe values and probability estimation state machine
+ */
+ sv = *st;
+ qe = jaritab[sv & 0x7F]; /* => Qe_Value */
+ nl = qe & 0xFF; qe >>= 8; /* Next_Index_LPS + Switch_MPS */
+ nm = qe & 0xFF; qe >>= 8; /* Next_Index_MPS */
+
+ /* Encode & estimation procedures per sections D.1.4 & D.1.5 */
+ e->a -= qe;
+ if (val != (sv >> 7)) {
+ /* Encode the less probable symbol */
+ if (e->a >= qe) {
+ /* If the interval size (qe) for the less probable symbol (LPS)
+ * is larger than the interval size for the MPS, then exchange
+ * the two symbols for coding efficiency, otherwise code the LPS
+ * as usual: */
+ e->c += e->a;
+ e->a = qe;
+ }
+ *st = (sv & 0x80) ^ nl; /* Estimate_after_LPS */
+ } else {
+ /* Encode the more probable symbol */
+ if (e->a >= 0x8000L)
+ return; /* A >= 0x8000 -> ready, no renormalization required */
+ if (e->a < qe) {
+ /* If the interval size (qe) for the less probable symbol (LPS)
+ * is larger than the interval size for the MPS, then exchange
+ * the two symbols for coding efficiency: */
+ e->c += e->a;
+ e->a = qe;
+ }
+ *st = (sv & 0x80) ^ nm; /* Estimate_after_MPS */
+ }
+
+ /* Renormalization & data output per section D.1.6 */
+ do {
+ e->a <<= 1;
+ e->c <<= 1;
+ if (--e->ct == 0) {
+ /* Another byte is ready for output */
+ temp = e->c >> 19;
+ if (temp > 0xFF) {
+ /* Handle overflow over all stacked 0xFF bytes */
+ if (e->buffer >= 0) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ emit_byte(e->buffer + 1, cinfo);
+ if (e->buffer + 1 == 0xFF)
+ emit_byte(0x00, cinfo);
+ }
+ e->zc += e->sc; /* carry-over converts stacked 0xFF bytes to 0x00 */
+ e->sc = 0;
+ /* Note: The 3 spacer bits in the C register guarantee
+ * that the new buffer byte can't be 0xFF here
+ * (see page 160 in the P&M JPEG book). */
+ e->buffer = temp & 0xFF; /* new output byte, might overflow later */
+ } else if (temp == 0xFF) {
+ ++e->sc; /* stack 0xFF byte (which might overflow later) */
+ } else {
+ /* Output all stacked 0xFF bytes, they will not overflow any more */
+ if (e->buffer == 0)
+ ++e->zc;
+ else if (e->buffer >= 0) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ emit_byte(e->buffer, cinfo);
+ }
+ if (e->sc) {
+ if (e->zc)
+ do emit_byte(0x00, cinfo);
+ while (--e->zc);
+ do {
+ emit_byte(0xFF, cinfo);
+ emit_byte(0x00, cinfo);
+ } while (--e->sc);
+ }
+ e->buffer = temp & 0xFF; /* new output byte (can still overflow) */
+ }
+ e->c &= 0x7FFFFL;
+ e->ct += 8;
+ }
+ } while (e->a < 0x8000L);
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(void)
+emit_restart (j_compress_ptr cinfo, int restart_num)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ int ci;
+ jpeg_component_info * compptr;
+
+ finish_pass(cinfo);
+
+ emit_byte(0xFF, cinfo);
+ emit_byte(JPEG_RST0 + restart_num, cinfo);
+
+ for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+ compptr = cinfo->cur_comp_info[ci];
+ /* Re-initialize statistics areas */
+ if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+ MEMZERO(entropy->dc_stats[compptr->dc_tbl_no], DC_STAT_BINS);
+ /* Reset DC predictions to 0 */
+ entropy->last_dc_val[ci] = 0;
+ entropy->dc_context[ci] = 0;
+ }
+ if (cinfo->progressive_mode == 0 || cinfo->Ss) {
+ MEMZERO(entropy->ac_stats[compptr->ac_tbl_no], AC_STAT_BINS);
+ }
+ }
+
+ /* Reset arithmetic encoding variables */
+ entropy->c = 0;
+ entropy->a = 0x10000L;
+ entropy->sc = 0;
+ entropy->zc = 0;
+ entropy->ct = 11;
+ entropy->buffer = -1; /* empty */
+}
+
+
+/*
+ * MCU encoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ unsigned char *st;
+ int blkn, ci, tbl;
+ int v, v2, m;
+ ISHIFT_TEMPS
+
+ /* Emit restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0) {
+ emit_restart(cinfo, entropy->next_restart_num);
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num++;
+ entropy->next_restart_num &= 7;
+ }
+ entropy->restarts_to_go--;
+ }
+
+ /* Encode the MCU data blocks */
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ block = MCU_data[blkn];
+ ci = cinfo->MCU_membership[blkn];
+ tbl = cinfo->cur_comp_info[ci]->dc_tbl_no;
+
+ /* Compute the DC value after the required point transform by Al.
+ * This is simply an arithmetic right shift.
+ */
+ m = IRIGHT_SHIFT((int) ((*block)[0]), cinfo->Al);
+
+ /* Sections F.1.4.1 & F.1.4.4.1: Encoding of DC coefficients */
+
+ /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+ st = entropy->dc_stats[tbl] + entropy->dc_context[ci];
+
+ /* Figure F.4: Encode_DC_DIFF */
+ if ((v = m - entropy->last_dc_val[ci]) == 0) {
+ arith_encode(cinfo, st, 0);
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ } else {
+ entropy->last_dc_val[ci] = m;
+ arith_encode(cinfo, st, 1);
+ /* Figure F.6: Encoding nonzero value v */
+ /* Figure F.7: Encoding the sign of v */
+ if (v > 0) {
+ arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+ st += 2; /* Table F.4: SP = S0 + 2 */
+ entropy->dc_context[ci] = 4; /* small positive diff category */
+ } else {
+ v = -v;
+ arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+ st += 3; /* Table F.4: SN = S0 + 3 */
+ entropy->dc_context[ci] = 8; /* small negative diff category */
+ }
+ /* Figure F.8: Encoding the magnitude category of v */
+ m = 0;
+ if (v -= 1) {
+ arith_encode(cinfo, st, 1);
+ m = 1;
+ v2 = v;
+ st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+ while (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st += 1;
+ }
+ }
+ arith_encode(cinfo, st, 0);
+ /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+ if (m < (int) (((INT32) 1 << cinfo->arith_dc_L[tbl]) >> 1))
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ else if (m > (int) (((INT32) 1 << cinfo->arith_dc_U[tbl]) >> 1))
+ entropy->dc_context[ci] += 8; /* large diff category */
+ /* Figure F.9: Encoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ arith_encode(cinfo, st, (m & v) ? 1 : 0);
+ }
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ unsigned char *st;
+ int tbl, k, ke;
+ int v, v2, m;
+
+ /* Emit restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0) {
+ emit_restart(cinfo, entropy->next_restart_num);
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num++;
+ entropy->next_restart_num &= 7;
+ }
+ entropy->restarts_to_go--;
+ }
+
+ /* Encode the MCU data block */
+ block = MCU_data[0];
+ tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
+
+ /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
+
+ /* Establish EOB (end-of-block) index */
+ for (ke = cinfo->Se + 1; ke > 1; ke--)
+ /* We must apply the point transform by Al. For AC coefficients this
+ * is an integer division with rounding towards 0. To do this portably
+ * in C, we shift after obtaining the absolute value.
+ */
+ if ((v = (*block)[jpeg_natural_order[ke - 1]]) >= 0) {
+ if (v >>= cinfo->Al) break;
+ } else {
+ v = -v;
+ if (v >>= cinfo->Al) break;
+ }
+
+ /* Figure F.5: Encode_AC_Coefficients */
+ for (k = cinfo->Ss; k < ke; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ arith_encode(cinfo, st, 0); /* EOB decision */
+ entropy->ac_stats[tbl][245] = 0;
+ for (;;) {
+ if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
+ if (v >>= cinfo->Al) {
+ arith_encode(cinfo, st + 1, 1);
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 0);
+ break;
+ }
+ } else {
+ v = -v;
+ if (v >>= cinfo->Al) {
+ arith_encode(cinfo, st + 1, 1);
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 1);
+ break;
+ }
+ }
+ arith_encode(cinfo, st + 1, 0); st += 3; k++;
+ }
+ st += 2;
+ /* Figure F.8: Encoding the magnitude category of v */
+ m = 0;
+ if (v -= 1) {
+ arith_encode(cinfo, st, 1);
+ m = 1;
+ v2 = v;
+ if (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st = entropy->ac_stats[tbl] +
+ (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+ while (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st += 1;
+ }
+ }
+ }
+ arith_encode(cinfo, st, 0);
+ /* Figure F.9: Encoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ arith_encode(cinfo, st, (m & v) ? 1 : 0);
+ }
+ /* Encode EOB decision only if k <= cinfo->Se */
+ if (k <= cinfo->Se) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ arith_encode(cinfo, st, 1);
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU encoding for DC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ unsigned char st[4];
+ int Al, blkn;
+
+ /* Emit restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0) {
+ emit_restart(cinfo, entropy->next_restart_num);
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num++;
+ entropy->next_restart_num &= 7;
+ }
+ entropy->restarts_to_go--;
+ }
+
+ Al = cinfo->Al;
+
+ /* Encode the MCU data blocks */
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ st[0] = 0; /* use fixed probability estimation */
+ /* We simply emit the Al'th bit of the DC coefficient value. */
+ arith_encode(cinfo, st, (MCU_data[blkn][0][0] >> Al) & 1);
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ unsigned char *st;
+ int tbl, k, ke, kex;
+ int v;
+
+ /* Emit restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0) {
+ emit_restart(cinfo, entropy->next_restart_num);
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num++;
+ entropy->next_restart_num &= 7;
+ }
+ entropy->restarts_to_go--;
+ }
+
+ /* Encode the MCU data block */
+ block = MCU_data[0];
+ tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
+
+ /* Section G.1.3.3: Encoding of AC coefficients */
+
+ /* Establish EOB (end-of-block) index */
+ for (ke = cinfo->Se + 1; ke > 1; ke--)
+ /* We must apply the point transform by Al. For AC coefficients this
+ * is an integer division with rounding towards 0. To do this portably
+ * in C, we shift after obtaining the absolute value.
+ */
+ if ((v = (*block)[jpeg_natural_order[ke - 1]]) >= 0) {
+ if (v >>= cinfo->Al) break;
+ } else {
+ v = -v;
+ if (v >>= cinfo->Al) break;
+ }
+
+ /* Establish EOBx (previous stage end-of-block) index */
+ for (kex = ke; kex > 1; kex--)
+ if ((v = (*block)[jpeg_natural_order[kex - 1]]) >= 0) {
+ if (v >>= cinfo->Ah) break;
+ } else {
+ v = -v;
+ if (v >>= cinfo->Ah) break;
+ }
+
+ /* Figure G.10: Encode_AC_Coefficients_SA */
+ for (k = cinfo->Ss; k < ke; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ if (k >= kex)
+ arith_encode(cinfo, st, 0); /* EOB decision */
+ entropy->ac_stats[tbl][245] = 0;
+ for (;;) {
+ if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
+ if (v >>= cinfo->Al) {
+ if (v >> 1) /* previously nonzero coef */
+ arith_encode(cinfo, st + 2, (v & 1));
+ else { /* newly nonzero coef */
+ arith_encode(cinfo, st + 1, 1);
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 0);
+ }
+ break;
+ }
+ } else {
+ v = -v;
+ if (v >>= cinfo->Al) {
+ if (v >> 1) /* previously nonzero coef */
+ arith_encode(cinfo, st + 2, (v & 1));
+ else { /* newly nonzero coef */
+ arith_encode(cinfo, st + 1, 1);
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 1);
+ }
+ break;
+ }
+ }
+ arith_encode(cinfo, st + 1, 0); st += 3; k++;
+ }
+ }
+ /* Encode EOB decision only if k <= cinfo->Se */
+ if (k <= cinfo->Se) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ arith_encode(cinfo, st, 1);
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * Encode and output one MCU's worth of arithmetic-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+encode_mcu (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ jpeg_component_info * compptr;
+ JBLOCKROW block;
+ unsigned char *st;
+ int blkn, ci, tbl, k, ke;
+ int v, v2, m;
+
+ /* Emit restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0) {
+ emit_restart(cinfo, entropy->next_restart_num);
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num++;
+ entropy->next_restart_num &= 7;
+ }
+ entropy->restarts_to_go--;
+ }
+
+ /* Encode the MCU data blocks */
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ block = MCU_data[blkn];
+ ci = cinfo->MCU_membership[blkn];
+ compptr = cinfo->cur_comp_info[ci];
+
+ /* Sections F.1.4.1 & F.1.4.4.1: Encoding of DC coefficients */
+
+ tbl = compptr->dc_tbl_no;
+
+ /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+ st = entropy->dc_stats[tbl] + entropy->dc_context[ci];
+
+ /* Figure F.4: Encode_DC_DIFF */
+ if ((v = (*block)[0] - entropy->last_dc_val[ci]) == 0) {
+ arith_encode(cinfo, st, 0);
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ } else {
+ entropy->last_dc_val[ci] = (*block)[0];
+ arith_encode(cinfo, st, 1);
+ /* Figure F.6: Encoding nonzero value v */
+ /* Figure F.7: Encoding the sign of v */
+ if (v > 0) {
+ arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+ st += 2; /* Table F.4: SP = S0 + 2 */
+ entropy->dc_context[ci] = 4; /* small positive diff category */
+ } else {
+ v = -v;
+ arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+ st += 3; /* Table F.4: SN = S0 + 3 */
+ entropy->dc_context[ci] = 8; /* small negative diff category */
+ }
+ /* Figure F.8: Encoding the magnitude category of v */
+ m = 0;
+ if (v -= 1) {
+ arith_encode(cinfo, st, 1);
+ m = 1;
+ v2 = v;
+ st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+ while (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st += 1;
+ }
+ }
+ arith_encode(cinfo, st, 0);
+ /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+ if (m < (int) (((INT32) 1 << cinfo->arith_dc_L[tbl]) >> 1))
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ else if (m > (int) (((INT32) 1 << cinfo->arith_dc_U[tbl]) >> 1))
+ entropy->dc_context[ci] += 8; /* large diff category */
+ /* Figure F.9: Encoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ arith_encode(cinfo, st, (m & v) ? 1 : 0);
+ }
+
+ /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
+
+ tbl = compptr->ac_tbl_no;
+
+ /* Establish EOB (end-of-block) index */
+ for (ke = DCTSIZE2; ke > 1; ke--)
+ if ((*block)[jpeg_natural_order[ke - 1]]) break;
+
+ /* Figure F.5: Encode_AC_Coefficients */
+ for (k = 1; k < ke; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ arith_encode(cinfo, st, 0); /* EOB decision */
+ while ((v = (*block)[jpeg_natural_order[k]]) == 0) {
+ arith_encode(cinfo, st + 1, 0); st += 3; k++;
+ }
+ arith_encode(cinfo, st + 1, 1);
+ /* Figure F.6: Encoding nonzero value v */
+ /* Figure F.7: Encoding the sign of v */
+ entropy->ac_stats[tbl][245] = 0;
+ if (v > 0) {
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 0);
+ } else {
+ v = -v;
+ arith_encode(cinfo, entropy->ac_stats[tbl] + 245, 1);
+ }
+ st += 2;
+ /* Figure F.8: Encoding the magnitude category of v */
+ m = 0;
+ if (v -= 1) {
+ arith_encode(cinfo, st, 1);
+ m = 1;
+ v2 = v;
+ if (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st = entropy->ac_stats[tbl] +
+ (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+ while (v2 >>= 1) {
+ arith_encode(cinfo, st, 1);
+ m <<= 1;
+ st += 1;
+ }
+ }
+ }
+ arith_encode(cinfo, st, 0);
+ /* Figure F.9: Encoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ arith_encode(cinfo, st, (m & v) ? 1 : 0);
+ }
+ /* Encode EOB decision only if k < DCTSIZE2 */
+ if (k < DCTSIZE2) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ arith_encode(cinfo, st, 1);
+ }
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * Initialize for an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass (j_compress_ptr cinfo, boolean gather_statistics)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ int ci, tbl;
+ jpeg_component_info * compptr;
+
+ if (gather_statistics)
+ /* Make sure to avoid that in the master control logic!
+ * We are fully adaptive here and need no extra
+ * statistics gathering pass!
+ */
+ ERREXIT(cinfo, JERR_NOT_COMPILED);
+
+ /* We assume jcmaster.c already validated the progressive scan parameters. */
+
+ /* Select execution routines */
+ if (cinfo->progressive_mode) {
+ if (cinfo->Ah == 0) {
+ if (cinfo->Ss == 0)
+ entropy->pub.encode_mcu = encode_mcu_DC_first;
+ else
+ entropy->pub.encode_mcu = encode_mcu_AC_first;
+ } else {
+ if (cinfo->Ss == 0)
+ entropy->pub.encode_mcu = encode_mcu_DC_refine;
+ else
+ entropy->pub.encode_mcu = encode_mcu_AC_refine;
+ }
+ } else
+ entropy->pub.encode_mcu = encode_mcu;
+
+ for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+ compptr = cinfo->cur_comp_info[ci];
+ /* Allocate & initialize requested statistics areas */
+ if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+ tbl = compptr->dc_tbl_no;
+ if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+ ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+ if (entropy->dc_stats[tbl] == NULL)
+ entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+ ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+ MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
+ /* Initialize DC predictions to 0 */
+ entropy->last_dc_val[ci] = 0;
+ entropy->dc_context[ci] = 0;
+ }
+ if (cinfo->progressive_mode == 0 || cinfo->Ss) {
+ tbl = compptr->ac_tbl_no;
+ if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+ ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+ if (entropy->ac_stats[tbl] == NULL)
+ entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+ ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+ MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
+#ifdef CALCULATE_SPECTRAL_CONDITIONING
+ if (cinfo->progressive_mode)
+ /* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
+ cinfo->arith_ac_K[tbl] = cinfo->Ss + ((8 + cinfo->Se - cinfo->Ss) >> 4);
+#endif
+ }
+ }
+
+ /* Initialize arithmetic encoding variables */
+ entropy->c = 0;
+ entropy->a = 0x10000L;
+ entropy->sc = 0;
+ entropy->zc = 0;
+ entropy->ct = 11;
+ entropy->buffer = -1; /* empty */
+
+ /* Initialize restart stuff */
+ entropy->restarts_to_go = cinfo->restart_interval;
+ entropy->next_restart_num = 0;
+}
+
+
+/*
+ * Module initialization routine for arithmetic entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_arith_encoder (j_compress_ptr cinfo)
+{
+ arith_entropy_ptr entropy;
+ int i;
+
+ entropy = (arith_entropy_ptr)
+ (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+ SIZEOF(arith_entropy_encoder));
+ cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
+ entropy->pub.start_pass = start_pass;
+ entropy->pub.finish_pass = finish_pass;
+
+ /* Mark tables unallocated */
+ for (i = 0; i < NUM_ARITH_TBLS; i++) {
+ entropy->dc_stats[i] = NULL;
+ entropy->ac_stats[i] = NULL;
+ }
+}
diff --git a/jcinit.c b/jcinit.c
index 5efffe3..a0f3e4f 100644
--- a/jcinit.c
+++ b/jcinit.c
@@ -41,9 +41,9 @@
/* Forward DCT */
jinit_forward_dct(cinfo);
/* Entropy encoding: either Huffman or arithmetic coding. */
- if (cinfo->arith_code) {
- ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
- } else {
+ if (cinfo->arith_code)
+ jinit_arith_encoder(cinfo);
+ else {
if (cinfo->progressive_mode) {
#ifdef C_PROGRESSIVE_SUPPORTED
jinit_phuff_encoder(cinfo);
diff --git a/jcmarker.c b/jcmarker.c
index 3d1e6c6..2d7bdc3 100644
--- a/jcmarker.c
+++ b/jcmarker.c
@@ -529,7 +529,10 @@
/* Emit the proper SOF marker */
if (cinfo->arith_code) {
- emit_sof(cinfo, M_SOF9); /* SOF code for arithmetic coding */
+ if (cinfo->progressive_mode)
+ emit_sof(cinfo, M_SOF10); /* SOF code for progressive arithmetic */
+ else
+ emit_sof(cinfo, M_SOF9); /* SOF code for sequential arithmetic */
} else {
if (cinfo->progressive_mode)
emit_sof(cinfo, M_SOF2); /* SOF code for progressive Huffman */
diff --git a/jcmaster.c b/jcmaster.c
index aab4020..366f7c0 100644
--- a/jcmaster.c
+++ b/jcmaster.c
@@ -433,7 +433,7 @@
/* Do Huffman optimization for a scan after the first one. */
select_scan_parameters(cinfo);
per_scan_setup(cinfo);
- if (cinfo->Ss != 0 || cinfo->Ah == 0 || cinfo->arith_code) {
+ if (cinfo->Ss != 0 || cinfo->Ah == 0) {
(*cinfo->entropy->start_pass) (cinfo, TRUE);
(*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
master->pub.call_pass_startup = FALSE;
@@ -567,7 +567,7 @@
cinfo->num_scans = 1;
}
- if (cinfo->progressive_mode) /* TEMPORARY HACK ??? */
+ if (cinfo->progressive_mode && cinfo->arith_code == 0) /* TEMPORARY HACK ??? */
cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
/* Initialize my private state */
diff --git a/jctrans.c b/jctrans.c
index 0e6d707..9753a43 100644
--- a/jctrans.c
+++ b/jctrans.c
@@ -167,7 +167,7 @@
/* Entropy encoding: either Huffman or arithmetic coding. */
if (cinfo->arith_code) {
- ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+ jinit_arith_encoder(cinfo);
} else {
if (cinfo->progressive_mode) {
#ifdef C_PROGRESSIVE_SUPPORTED
diff --git a/jdarith.c b/jdarith.c
new file mode 100644
index 0000000..1ef513e
--- /dev/null
+++ b/jdarith.c
@@ -0,0 +1,762 @@
+/*
+ * jdarith.c
+ *
+ * Copyright (C) 1997, Guido Vollbeding <guivol@esc.de>.
+ * This file is NOT part of the Independent JPEG Group's software
+ * for legal reasons.
+ * See the accompanying README file for conditions of distribution and use.
+ *
+ * This file contains portable arithmetic entropy decoding routines for JPEG
+ * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
+ *
+ * Both sequential and progressive modes are supported in this single module.
+ *
+ * Suspension is not currently supported in this module.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Expanded entropy decoder object for arithmetic decoding. */
+
+typedef struct {
+ struct jpeg_entropy_decoder pub; /* public fields */
+
+ INT32 c; /* C register, base of coding interval + input bit buffer */
+ INT32 a; /* A register, normalized size of coding interval */
+ int ct; /* bit shift counter, # of bits left in bit buffer part of C */
+ /* init: ct = -16 */
+ /* run: ct = 0..7 */
+ /* error: ct = -1 */
+ int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+ int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
+
+ unsigned int restarts_to_go; /* MCUs left in this restart interval */
+
+ /* Pointers to statistics areas (these workspaces have image lifespan) */
+ unsigned char * dc_stats[NUM_ARITH_TBLS];
+ unsigned char * ac_stats[NUM_ARITH_TBLS];
+} arith_entropy_decoder;
+
+typedef arith_entropy_decoder * arith_entropy_ptr;
+
+/* The following two definitions specify the allocation chunk size
+ * for the statistics area.
+ * According to sections F.1.4.4.1.3 and F.1.4.4.2, we need at least
+ * 49 statistics bins for DC, and 245 statistics bins for AC coding.
+ * Note that we use one additional AC bin for codings with fixed
+ * probability (0.5), thus the minimum number for AC is 246.
+ *
+ * We use a compact representation with 1 byte per statistics bin,
+ * thus the numbers directly represent byte sizes.
+ * This 1 byte per statistics bin contains the meaning of the MPS
+ * (more probable symbol) in the highest bit (mask 0x80), and the
+ * index into the probability estimation state machine table
+ * in the lower bits (mask 0x7F).
+ */
+
+#define DC_STAT_BINS 64
+#define AC_STAT_BINS 256
+
+
+LOCAL(int)
+get_byte (j_decompress_ptr cinfo)
+/* Read next input byte; we do not support suspension in this module. */
+{
+ struct jpeg_source_mgr * src = cinfo->src;
+
+ if (src->bytes_in_buffer == 0)
+ if (! (*src->fill_input_buffer) (cinfo))
+ ERREXIT(cinfo, JERR_CANT_SUSPEND);
+ src->bytes_in_buffer--;
+ return GETJOCTET(*src->next_input_byte++);
+}
+
+
+/*
+ * The core arithmetic decoding routine (common in JPEG and JBIG).
+ * This needs to go as fast as possible.
+ * Machine-dependent optimization facilities
+ * are not utilized in this portable implementation.
+ * However, this code should be fairly efficient and
+ * may be a good base for further optimizations anyway.
+ *
+ * Return value is 0 or 1 (binary decision).
+ *
+ * Note: I've changed the handling of the code base & bit
+ * buffer register C compared to other implementations
+ * based on the standards layout & procedures.
+ * While it also contains both the actual base of the
+ * coding interval (16 bits) and the next-bits buffer,
+ * the cut-point between these two parts is floating
+ * (instead of fixed) with the bit shift counter CT.
+ * Thus, we also need only one (variable instead of
+ * fixed size) shift for the LPS/MPS decision, and
+ * we can get away with any renormalization update
+ * of C (except for new data insertion, of course).
+ *
+ * I've also introduced a new scheme for accessing
+ * the probability estimation state machine table,
+ * derived from Markus Kuhn's JBIG implementation.
+ */
+
+LOCAL(int)
+arith_decode (j_decompress_ptr cinfo, unsigned char *st)
+{
+ extern const INT32 jaritab[];
+ register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
+ register unsigned char nl, nm;
+ register INT32 qe, temp;
+ register int sv, data;
+
+ /* Renormalization & data input per section D.2.6 */
+ while (e->a < 0x8000L) {
+ if (--e->ct < 0) {
+ /* Need to fetch next data byte */
+ if (cinfo->unread_marker)
+ data = 0; /* stuff zero data */
+ else {
+ data = get_byte(cinfo); /* read next input byte */
+ if (data == 0xFF) { /* zero stuff or marker code */
+ do data = get_byte(cinfo);
+ while (data == 0xFF); /* swallow extra 0xFF bytes */
+ if (data == 0)
+ data = 0xFF; /* discard stuffed zero byte */
+ else {
+ /* Note: Different from the Huffman decoder, hitting
+ * a marker while processing the compressed data
+ * segment is legal in arithmetic coding.
+ * The convention is to supply zero data
+ * then until decoding is complete.
+ */
+ cinfo->unread_marker = data;
+ data = 0;
+ }
+ }
+ }
+ e->c = (e->c << 8) | data; /* insert data into C register */
+ if ((e->ct += 8) < 0) /* update bit shift counter */
+ /* Need more initial bytes */
+ if (++e->ct == 0)
+ /* Got 2 initial bytes -> re-init A and exit loop */
+ e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
+ }
+ e->a <<= 1;
+ }
+
+ /* Fetch values from our compact representation of Table D.2:
+ * Qe values and probability estimation state machine
+ */
+ sv = *st;
+ qe = jaritab[sv & 0x7F]; /* => Qe_Value */
+ nl = qe & 0xFF; qe >>= 8; /* Next_Index_LPS + Switch_MPS */
+ nm = qe & 0xFF; qe >>= 8; /* Next_Index_MPS */
+
+ /* Decode & estimation procedures per sections D.2.4 & D.2.5 */
+ temp = e->a - qe;
+ e->a = temp;
+ temp <<= e->ct;
+ if (e->c >= temp) {
+ e->c -= temp;
+ /* Conditional LPS (less probable symbol) exchange */
+ if (e->a < qe) {
+ e->a = qe;
+ *st = (sv & 0x80) ^ nm; /* Estimate_after_MPS */
+ } else {
+ e->a = qe;
+ *st = (sv & 0x80) ^ nl; /* Estimate_after_LPS */
+ sv ^= 0x80; /* Exchange LPS/MPS */
+ }
+ } else if (e->a < 0x8000L) {
+ /* Conditional MPS (more probable symbol) exchange */
+ if (e->a < qe) {
+ *st = (sv & 0x80) ^ nl; /* Estimate_after_LPS */
+ sv ^= 0x80; /* Exchange LPS/MPS */
+ } else {
+ *st = (sv & 0x80) ^ nm; /* Estimate_after_MPS */
+ }
+ }
+
+ return sv >> 7;
+}
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ */
+
+LOCAL(void)
+process_restart (j_decompress_ptr cinfo)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ int ci;
+ jpeg_component_info * compptr;
+
+ /* Advance past the RSTn marker */
+ if (! (*cinfo->marker->read_restart_marker) (cinfo))
+ ERREXIT(cinfo, JERR_CANT_SUSPEND);
+
+ for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+ compptr = cinfo->cur_comp_info[ci];
+ /* Re-initialize statistics areas */
+ if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+ MEMZERO(entropy->dc_stats[compptr->dc_tbl_no], DC_STAT_BINS);
+ /* Reset DC predictions to 0 */
+ entropy->last_dc_val[ci] = 0;
+ entropy->dc_context[ci] = 0;
+ }
+ if (cinfo->progressive_mode == 0 || cinfo->Ss) {
+ MEMZERO(entropy->ac_stats[compptr->ac_tbl_no], AC_STAT_BINS);
+ }
+ }
+
+ /* Reset arithmetic decoding variables */
+ entropy->c = 0;
+ entropy->a = 0;
+ entropy->ct = -16; /* force reading 2 initial bytes to fill C */
+
+ /* Reset restart counter */
+ entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Arithmetic MCU decoding.
+ * Each of these routines decodes and returns one MCU's worth of
+ * arithmetic-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_data[i]. WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER.
+ */
+
+/*
+ * MCU decoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ unsigned char *st;
+ int blkn, ci, tbl, sign;
+ int v, m;
+
+ /* Process restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0)
+ process_restart(cinfo);
+ entropy->restarts_to_go--;
+ }
+
+ if (entropy->ct == -1) return TRUE; /* if error do nothing */
+
+ /* Outer loop handles each block in the MCU */
+
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ block = MCU_data[blkn];
+ ci = cinfo->MCU_membership[blkn];
+ tbl = cinfo->cur_comp_info[ci]->dc_tbl_no;
+
+ /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */
+
+ /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+ st = entropy->dc_stats[tbl] + entropy->dc_context[ci];
+
+ /* Figure F.19: Decode_DC_DIFF */
+ if (arith_decode(cinfo, st) == 0)
+ entropy->dc_context[ci] = 0;
+ else {
+ /* Figure F.21: Decoding nonzero value v */
+ /* Figure F.22: Decoding the sign of v */
+ sign = arith_decode(cinfo, st + 1);
+ st += 2; st += sign;
+ /* Figure F.23: Decoding the magnitude category of v */
+ if ((m = arith_decode(cinfo, st)) != 0) {
+ st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+ while (arith_decode(cinfo, st)) {
+ if ((m <<= 1) == 0x8000) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* magnitude overflow */
+ return TRUE;
+ }
+ st += 1;
+ }
+ }
+ /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+ if (m < (int) (((INT32) 1 << cinfo->arith_dc_L[tbl]) >> 1))
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ else if (m > (int) (((INT32) 1 << cinfo->arith_dc_U[tbl]) >> 1))
+ entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+ else
+ entropy->dc_context[ci] = 4 + (sign * 4); /* small diff category */
+ v = m;
+ /* Figure F.24: Decoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ if (arith_decode(cinfo, st)) v |= m;
+ v += 1; if (sign) v = -v;
+ entropy->last_dc_val[ci] += v;
+ }
+
+ /* Scale and output the DC coefficient (assumes jpeg_natural_order[0]=0) */
+ (*block)[0] = (JCOEF) (entropy->last_dc_val[ci] << cinfo->Al);
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ unsigned char *st;
+ int tbl, sign, k;
+ int v, m;
+
+ /* Process restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0)
+ process_restart(cinfo);
+ entropy->restarts_to_go--;
+ }
+
+ if (entropy->ct == -1) return TRUE; /* if error do nothing */
+
+ /* There is always only one block per MCU */
+ block = MCU_data[0];
+ tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
+
+ /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
+
+ /* Figure F.20: Decode_AC_coefficients */
+ for (k = cinfo->Ss; k <= cinfo->Se; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ if (arith_decode(cinfo, st)) break; /* EOB flag */
+ while (arith_decode(cinfo, st + 1) == 0) {
+ st += 3; k++;
+ if (k > cinfo->Se) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* spectral overflow */
+ return TRUE;
+ }
+ }
+ /* Figure F.21: Decoding nonzero value v */
+ /* Figure F.22: Decoding the sign of v */
+ entropy->ac_stats[tbl][245] = 0;
+ sign = arith_decode(cinfo, entropy->ac_stats[tbl] + 245);
+ st += 2;
+ /* Figure F.23: Decoding the magnitude category of v */
+ if ((m = arith_decode(cinfo, st)) != 0) {
+ if (arith_decode(cinfo, st)) {
+ m <<= 1;
+ st = entropy->ac_stats[tbl] +
+ (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+ while (arith_decode(cinfo, st)) {
+ if ((m <<= 1) == 0x8000) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* magnitude overflow */
+ return TRUE;
+ }
+ st += 1;
+ }
+ }
+ }
+ v = m;
+ /* Figure F.24: Decoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ if (arith_decode(cinfo, st)) v |= m;
+ v += 1; if (sign) v = -v;
+ /* Scale and output coefficient in natural (dezigzagged) order */
+ (*block)[jpeg_natural_order[k]] = (JCOEF) (v << cinfo->Al);
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU decoding for DC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ unsigned char st[4];
+ int p1, blkn;
+
+ /* Process restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0)
+ process_restart(cinfo);
+ entropy->restarts_to_go--;
+ }
+
+ p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */
+
+ /* Outer loop handles each block in the MCU */
+
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ st[0] = 0; /* use fixed probability estimation */
+ /* Encoded data is simply the next bit of the two's-complement DC value */
+ if (arith_decode(cinfo, st))
+ MCU_data[blkn][0][0] |= p1;
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ JBLOCKROW block;
+ JCOEFPTR thiscoef;
+ unsigned char *st;
+ int tbl, k, kex;
+ int p1, m1;
+
+ /* Process restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0)
+ process_restart(cinfo);
+ entropy->restarts_to_go--;
+ }
+
+ if (entropy->ct == -1) return TRUE; /* if error do nothing */
+
+ /* There is always only one block per MCU */
+ block = MCU_data[0];
+ tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
+
+ p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */
+ m1 = (-1) << cinfo->Al; /* -1 in the bit position being coded */
+
+ /* Establish EOBx (previous stage end-of-block) index */
+ for (kex = cinfo->Se + 1; kex > 1; kex--)
+ if ((*block)[jpeg_natural_order[kex - 1]]) break;
+
+ for (k = cinfo->Ss; k <= cinfo->Se; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ if (k >= kex)
+ if (arith_decode(cinfo, st)) break; /* EOB flag */
+ for (;;) {
+ thiscoef = *block + jpeg_natural_order[k];
+ if (*thiscoef) { /* previously nonzero coef */
+ if (arith_decode(cinfo, st + 2))
+ if (*thiscoef < 0)
+ *thiscoef += m1;
+ else
+ *thiscoef += p1;
+ break;
+ }
+ if (arith_decode(cinfo, st + 1)) { /* newly nonzero coef */
+ entropy->ac_stats[tbl][245] = 0;
+ if (arith_decode(cinfo, entropy->ac_stats[tbl] + 245))
+ *thiscoef = m1;
+ else
+ *thiscoef = p1;
+ break;
+ }
+ st += 3; k++;
+ if (k > cinfo->Se) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* spectral overflow */
+ return TRUE;
+ }
+ }
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * Decode one MCU's worth of arithmetic-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ jpeg_component_info * compptr;
+ JBLOCKROW block;
+ unsigned char *st;
+ int blkn, ci, tbl, sign, k;
+ int v, m;
+
+ /* Process restart marker if needed */
+ if (cinfo->restart_interval) {
+ if (entropy->restarts_to_go == 0)
+ process_restart(cinfo);
+ entropy->restarts_to_go--;
+ }
+
+ if (entropy->ct == -1) return TRUE; /* if error do nothing */
+
+ /* Outer loop handles each block in the MCU */
+
+ for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+ block = MCU_data[blkn];
+ ci = cinfo->MCU_membership[blkn];
+ compptr = cinfo->cur_comp_info[ci];
+
+ /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */
+
+ tbl = compptr->dc_tbl_no;
+
+ /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+ st = entropy->dc_stats[tbl] + entropy->dc_context[ci];
+
+ /* Figure F.19: Decode_DC_DIFF */
+ if (arith_decode(cinfo, st) == 0)
+ entropy->dc_context[ci] = 0;
+ else {
+ /* Figure F.21: Decoding nonzero value v */
+ /* Figure F.22: Decoding the sign of v */
+ sign = arith_decode(cinfo, st + 1);
+ st += 2; st += sign;
+ /* Figure F.23: Decoding the magnitude category of v */
+ if ((m = arith_decode(cinfo, st)) != 0) {
+ st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+ while (arith_decode(cinfo, st)) {
+ if ((m <<= 1) == 0x8000) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* magnitude overflow */
+ return TRUE;
+ }
+ st += 1;
+ }
+ }
+ /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+ if (m < (int) (((INT32) 1 << cinfo->arith_dc_L[tbl]) >> 1))
+ entropy->dc_context[ci] = 0; /* zero diff category */
+ else if (m > (int) (((INT32) 1 << cinfo->arith_dc_U[tbl]) >> 1))
+ entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+ else
+ entropy->dc_context[ci] = 4 + (sign * 4); /* small diff category */
+ v = m;
+ /* Figure F.24: Decoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ if (arith_decode(cinfo, st)) v |= m;
+ v += 1; if (sign) v = -v;
+ entropy->last_dc_val[ci] += v;
+ }
+
+ (*block)[0] = (JCOEF) entropy->last_dc_val[ci];
+
+ /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
+
+ tbl = compptr->ac_tbl_no;
+
+ /* Figure F.20: Decode_AC_coefficients */
+ for (k = 1; k < DCTSIZE2; k++) {
+ st = entropy->ac_stats[tbl] + 3 * (k - 1);
+ if (arith_decode(cinfo, st)) break; /* EOB flag */
+ while (arith_decode(cinfo, st + 1) == 0) {
+ st += 3; k++;
+ if (k >= DCTSIZE2) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* spectral overflow */
+ return TRUE;
+ }
+ }
+ /* Figure F.21: Decoding nonzero value v */
+ /* Figure F.22: Decoding the sign of v */
+ entropy->ac_stats[tbl][245] = 0;
+ sign = arith_decode(cinfo, entropy->ac_stats[tbl] + 245);
+ st += 2;
+ /* Figure F.23: Decoding the magnitude category of v */
+ if ((m = arith_decode(cinfo, st)) != 0) {
+ if (arith_decode(cinfo, st)) {
+ m <<= 1;
+ st = entropy->ac_stats[tbl] +
+ (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+ while (arith_decode(cinfo, st)) {
+ if ((m <<= 1) == 0x8000) {
+ WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+ entropy->ct = -1; /* magnitude overflow */
+ return TRUE;
+ }
+ st += 1;
+ }
+ }
+ }
+ v = m;
+ /* Figure F.24: Decoding the magnitude bit pattern of v */
+ st += 14;
+ while (m >>= 1)
+ if (arith_decode(cinfo, st)) v |= m;
+ v += 1; if (sign) v = -v;
+ (*block)[jpeg_natural_order[k]] = (JCOEF) v;
+ }
+ }
+
+ return TRUE;
+}
+
+
+/*
+ * Initialize for an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass (j_decompress_ptr cinfo)
+{
+ arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
+ int ci, tbl;
+ jpeg_component_info * compptr;
+
+ if (cinfo->progressive_mode) {
+ /* Validate progressive scan parameters */
+ if (cinfo->Ss == 0) {
+ if (cinfo->Se != 0)
+ goto bad;
+ } else {
+ /* need not check Ss/Se < 0 since they came from unsigned bytes */
+ if (cinfo->Se < cinfo->Ss || cinfo->Se >= DCTSIZE2)
+ goto bad;
+ /* AC scans may have only one component */
+ if (cinfo->comps_in_scan != 1)
+ goto bad;
+ }
+ if (cinfo->Ah != 0) {
+ /* Successive approximation refinement scan: must have Al = Ah-1. */
+ if (cinfo->Ah-1 != cinfo->Al)
+ goto bad;
+ }
+ if (cinfo->Al > 13) { /* need not check for < 0 */
+ bad:
+ ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
+ cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+ }
+ /* Update progression status, and verify that scan order is legal.
+ * Note that inter-scan inconsistencies are treated as warnings
+ * not fatal errors ... not clear if this is right way to behave.
+ */
+ for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+ int coefi, cindex = cinfo->cur_comp_info[ci]->component_index;
+ int *coef_bit_ptr = & cinfo->coef_bits[cindex][0];
+ if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
+ WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+ for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
+ int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
+ if (cinfo->Ah != expected)
+ WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+ coef_bit_ptr[coefi] = cinfo->Al;
+ }
+ }
+ /* Select MCU decoding routine */
+ if (cinfo->Ah == 0) {
+ if (cinfo->Ss == 0)
+ entropy->pub.decode_mcu = decode_mcu_DC_first;
+ else
+ entropy->pub.decode_mcu = decode_mcu_AC_first;
+ } else {
+ if (cinfo->Ss == 0)
+ entropy->pub.decode_mcu = decode_mcu_DC_refine;
+ else
+ entropy->pub.decode_mcu = decode_mcu_AC_refine;
+ }
+ } else {
+ /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
+ * This ought to be an error condition, but we make it a warning because
+ * there are some baseline files out there with all zeroes in these bytes.
+ */
+ if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2-1 ||
+ cinfo->Ah != 0 || cinfo->Al != 0)
+ WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
+ /* Select MCU decoding routine */
+ entropy->pub.decode_mcu = decode_mcu;
+ }
+
+ for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+ compptr = cinfo->cur_comp_info[ci];
+ /* Allocate & initialize requested statistics areas */
+ if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+ tbl = compptr->dc_tbl_no;
+ if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+ ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+ if (entropy->dc_stats[tbl] == NULL)
+ entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+ ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+ MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
+ /* Initialize DC predictions to 0 */
+ entropy->last_dc_val[ci] = 0;
+ entropy->dc_context[ci] = 0;
+ }
+ if (cinfo->progressive_mode == 0 || cinfo->Ss) {
+ tbl = compptr->ac_tbl_no;
+ if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+ ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+ if (entropy->ac_stats[tbl] == NULL)
+ entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+ ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+ MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
+ }
+ }
+
+ /* Initialize arithmetic decoding variables */
+ entropy->c = 0;
+ entropy->a = 0;
+ entropy->ct = -16; /* force reading 2 initial bytes to fill C */
+
+ /* Initialize restart counter */
+ entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Module initialization routine for arithmetic entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_arith_decoder (j_decompress_ptr cinfo)
+{
+ arith_entropy_ptr entropy;
+ int i;
+
+ entropy = (arith_entropy_ptr)
+ (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+ SIZEOF(arith_entropy_decoder));
+ cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
+ entropy->pub.start_pass = start_pass;
+
+ /* Mark tables unallocated */
+ for (i = 0; i < NUM_ARITH_TBLS; i++) {
+ entropy->dc_stats[i] = NULL;
+ entropy->ac_stats[i] = NULL;
+ }
+
+ if (cinfo->progressive_mode) {
+ /* Create progression status table */
+ int *coef_bit_ptr, ci;
+ cinfo->coef_bits = (int (*)[DCTSIZE2])
+ (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+ cinfo->num_components*DCTSIZE2*SIZEOF(int));
+ coef_bit_ptr = & cinfo->coef_bits[0][0];
+ for (ci = 0; ci < cinfo->num_components; ci++)
+ for (i = 0; i < DCTSIZE2; i++)
+ *coef_bit_ptr++ = -1;
+ }
+}
diff --git a/jdmaster.c b/jdmaster.c
index 2802c5b..6d23fee 100644
--- a/jdmaster.c
+++ b/jdmaster.c
@@ -373,7 +373,7 @@
jinit_inverse_dct(cinfo);
/* Entropy decoding: either Huffman or arithmetic coding. */
if (cinfo->arith_code) {
- ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+ jinit_arith_decoder(cinfo);
} else {
if (cinfo->progressive_mode) {
#ifdef D_PROGRESSIVE_SUPPORTED
diff --git a/jdtrans.c b/jdtrans.c
index 6c0ab71..7a3b268 100644
--- a/jdtrans.c
+++ b/jdtrans.c
@@ -100,9 +100,9 @@
cinfo->buffered_image = TRUE;
/* Entropy decoding: either Huffman or arithmetic coding. */
- if (cinfo->arith_code) {
- ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
- } else {
+ if (cinfo->arith_code)
+ jinit_arith_decoder(cinfo);
+ else {
if (cinfo->progressive_mode) {
#ifdef D_PROGRESSIVE_SUPPORTED
jinit_phuff_decoder(cinfo);
diff --git a/jerror.h b/jerror.h
index fc2fffe..109e3d3 100644
--- a/jerror.h
+++ b/jerror.h
@@ -93,6 +93,7 @@
JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
JMESSAGE(JERR_NOT_COMPILED, "Requested feature was omitted at compile time")
+JMESSAGE(JERR_NO_ARITH_TABLE, "Arithmetic table 0x%02x was not defined")
JMESSAGE(JERR_NO_BACKING_STORE, "Backing store not supported")
JMESSAGE(JERR_NO_HUFF_TABLE, "Huffman table 0x%02x was not defined")
JMESSAGE(JERR_NO_IMAGE, "JPEG datastream contains no image")
@@ -170,6 +171,7 @@
JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
+JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
JMESSAGE(JWRN_BOGUS_PROGRESSION,
"Inconsistent progression sequence for component %d coefficient %d")
JMESSAGE(JWRN_EXTRANEOUS_DATA,
diff --git a/jmorecfg.h b/jmorecfg.h
index 54a7d1c..bd6d605 100644
--- a/jmorecfg.h
+++ b/jmorecfg.h
@@ -266,7 +266,7 @@
/* Encoder capability options: */
-#undef C_ARITH_CODING_SUPPORTED /* Arithmetic coding back end? */
+#define C_ARITH_CODING_SUPPORTED /* Arithmetic coding back end? */
#define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
#define C_PROGRESSIVE_SUPPORTED /* Progressive JPEG? (Requires MULTISCAN)*/
#define ENTROPY_OPT_SUPPORTED /* Optimization of entropy coding parms? */
@@ -282,7 +282,7 @@
/* Decoder capability options: */
-#undef D_ARITH_CODING_SUPPORTED /* Arithmetic coding back end? */
+#define D_ARITH_CODING_SUPPORTED /* Arithmetic coding back end? */
#define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
#define D_PROGRESSIVE_SUPPORTED /* Progressive JPEG? (Requires MULTISCAN)*/
#define SAVE_MARKERS_SUPPORTED /* jpeg_save_markers() needed? */
diff --git a/jpegint.h b/jpegint.h
index 95b00d4..3378133 100644
--- a/jpegint.h
+++ b/jpegint.h
@@ -345,6 +345,7 @@
EXTERN(void) jinit_forward_dct JPP((j_compress_ptr cinfo));
EXTERN(void) jinit_huff_encoder JPP((j_compress_ptr cinfo));
EXTERN(void) jinit_phuff_encoder JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_arith_encoder JPP((j_compress_ptr cinfo));
EXTERN(void) jinit_marker_writer JPP((j_compress_ptr cinfo));
/* Decompression module initialization routines */
EXTERN(void) jinit_master_decompress JPP((j_decompress_ptr cinfo));
@@ -358,6 +359,7 @@
EXTERN(void) jinit_marker_reader JPP((j_decompress_ptr cinfo));
EXTERN(void) jinit_huff_decoder JPP((j_decompress_ptr cinfo));
EXTERN(void) jinit_phuff_decoder JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_arith_decoder JPP((j_decompress_ptr cinfo));
EXTERN(void) jinit_inverse_dct JPP((j_decompress_ptr cinfo));
EXTERN(void) jinit_upsampler JPP((j_decompress_ptr cinfo));
EXTERN(void) jinit_color_deconverter JPP((j_decompress_ptr cinfo));
diff --git a/makefile.cfg b/makefile.cfg
index f25e42e..7968319 100644
--- a/makefile.cfg
+++ b/makefile.cfg
@@ -80,7 +80,7 @@
jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
- jquant2.c jutils.c jmemmgr.c
+ jquant2.c jutils.c jmemmgr.c jaricom.c jcarith.c jdarith.c
# memmgr back ends: compile only one of these into a working library
SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
@@ -110,19 +110,19 @@
DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
$(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
# library object files common to compression and decompression
-COMOBJECTS= jcomapi.$(O) jutils.$(O) jerror.$(O) jmemmgr.$(O) $(SYSDEPMEM)
+COMOBJECTS= jcomapi.$(O) jutils.$(O) jerror.$(O) jmemmgr.$(O) jaricom.$(O) $(SYSDEPMEM)
# compression library object files
CLIBOBJECTS= jcapimin.$(O) jcapistd.$(O) jctrans.$(O) jcparam.$(O) \
jdatadst.$(O) jcinit.$(O) jcmaster.$(O) jcmarker.$(O) jcmainct.$(O) \
jcprepct.$(O) jccoefct.$(O) jccolor.$(O) jcsample.$(O) jchuff.$(O) \
jcphuff.$(O) jcdctmgr.$(O) jfdctfst.$(O) jfdctflt.$(O) \
- jfdctint.$(O)
+ jfdctint.$(O) jcarith.$(O)
# decompression library object files
DLIBOBJECTS= jdapimin.$(O) jdapistd.$(O) jdtrans.$(O) jdatasrc.$(O) \
jdmaster.$(O) jdinput.$(O) jdmarker.$(O) jdhuff.$(O) jdphuff.$(O) \
jdmainct.$(O) jdcoefct.$(O) jdpostct.$(O) jddctmgr.$(O) \
jidctfst.$(O) jidctflt.$(O) jidctint.$(O) jidctred.$(O) \
- jdsample.$(O) jdcolor.$(O) jquant1.$(O) jquant2.$(O) jdmerge.$(O)
+ jdsample.$(O) jdcolor.$(O) jquant1.$(O) jquant2.$(O) jdmerge.$(O) jdarith.$(O)
# These objectfiles are included in libjpeg.a
LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
# object files for sample applications (excluding library files)
@@ -317,3 +317,6 @@
wrbmp.$(O): wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
rdrle.$(O): rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
wrrle.$(O): wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+jcarith.$(O): jcarith.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdarith.$(O): jdarith.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jaricom.$(O): jaricom.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
