| 1 |
History of the reverse engineering of the ax203 compression |
| 2 |
----------------------------------------------------------- |
| 3 |
|
| 4 |
There are 4 (known) different compression algorithms: |
| 5 |
|
| 6 |
YUV This algorithm uses a simple YUV format, with the UV members |
| 7 |
subsampled at a 2x2 resolution. It uses 5 bits for the Y |
| 8 |
values and 6 bits for the U and V values, cramming 4 pixels |
| 9 |
into 4 bytes. This is used with firmware v3.3.x (usb-id |
| 10 |
1908:1315) and v3.4.x frames (usb-id 1908:1320) |
| 11 |
|
| 12 |
YUV-DELTA This version uses a less simple YUV format, it still |
| 13 |
subsamples UV at a 2x2 resolution. It then further uses a |
| 14 |
delta compression to store 4 values into 2 bytes. Data is |
| 15 |
stored in to 4x4 pixels block, Using 6 x 2 bytes, 4 x 4 Y |
| 16 |
pixels + 4 U and 4 V pixels. This is used with firmware v3.3.x |
| 17 |
(usb-id 1908:1315) and v3.4.x frames (usb-id 1908:1320) |
| 18 |
|
| 19 |
JPEG (ax3003) Regular JPEG |
| 20 |
|
| 21 |
JPEG (ax206) Uses a JPEG derived format reverse engineering this one |
| 22 |
was a challenge, see below for the full story. This is used with |
| 23 |
ax206 based firmware v3.5.x frames (usb-id 1908:0102) |
| 24 |
|
| 25 |
After some failed attempts by me (Hans de Goede), I asked help |
| 26 |
from Bertrik Sikken with reverse engineering the JPEG compression. |
| 27 |
He went the route of disassembling the windows binaries for uploading |
| 28 |
pictures, and wrote an algorithm description of the decompression on |
| 29 |
the picframe wiki: |
| 30 |
http://picframe.spritesserver.nl/wiki/index.php/ImageEncodingAx206 |
| 31 |
|
| 32 |
A copy of this page is included below. |
| 33 |
|
| 34 |
I used this description to write decompression code. For the decompression |
| 35 |
code I started with tinyjpeg, which with I'm familiar from my libv4l |
| 36 |
work. The ax203 camlib contains a modified tinyjpeg copy which was |
| 37 |
modified to work with ax203's JPEG flavor. Tinyjpeg is: |
| 38 |
|
| 39 |
/* |
| 40 |
* Small jpeg decoder library |
| 41 |
* |
| 42 |
* Copyright (c) 2006, Luc Saillard <[email protected]> |
| 43 |
* All rights reserved. |
| 44 |
* Redistribution and use in source and binary forms, with or without |
| 45 |
* modification, are permitted provided that the following conditions are met: |
| 46 |
* |
| 47 |
* - Redistributions of source code must retain the above copyright notice, |
| 48 |
* this list of conditions and the following disclaimer. |
| 49 |
* |
| 50 |
* - Redistributions in binary form must reproduce the above copyright notice, |
| 51 |
* this list of conditions and the following disclaimer in the documentation |
| 52 |
* and/or other materials provided with the distribution. |
| 53 |
* |
| 54 |
* - Neither the name of the author nor the names of its contributors may be |
| 55 |
* used to endorse or promote products derived from this software without |
| 56 |
* specific prior written permission. |
| 57 |
* |
| 58 |
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 59 |
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 60 |
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 61 |
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
| 62 |
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
| 63 |
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
| 64 |
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
| 65 |
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
| 66 |
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| 67 |
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| 68 |
* POSSIBILITY OF SUCH DAMAGE. |
| 69 |
* |
| 70 |
*/ |
| 71 |
|
| 72 |
Note that tinyjpeg includes jidctflt.c from libjpeg, see below for |
| 73 |
libjpeg's copyright info. |
| 74 |
|
| 75 |
Once I had working decompression code I needed to write compression code, |
| 76 |
as tinyjpeg does not do compression I turned to libjpeg. I did not want |
| 77 |
to include a modified libjpeg and luckily with some tricks it was not needed |
| 78 |
to include libjpeg code other then jpeg_memsrcdest.c, which is not part of |
| 79 |
the standard libjpeg distribution. This code is: |
| 80 |
|
| 81 |
/* |
| 82 |
* jpeg_memsrcdest.c and jidctflt.c |
| 83 |
* |
| 84 |
* Copyright (C) 1994-1998, Thomas G. Lane. |
| 85 |
* This file is part of the Independent JPEG Group's software. |
| 86 |
* |
| 87 |
* The authors make NO WARRANTY or representation, either express or implied, |
| 88 |
* with respect to this software, its quality, accuracy, merchantability, or |
| 89 |
* fitness for a particular purpose. This software is provided "AS IS", and you, |
| 90 |
* its user, assume the entire risk as to its quality and accuracy. |
| 91 |
* |
| 92 |
* This software is copyright (C) 1991-1998, Thomas G. Lane. |
| 93 |
* All Rights Reserved except as specified below. |
| 94 |
* |
| 95 |
* Permission is hereby granted to use, copy, modify, and distribute this |
| 96 |
* software (or portions thereof) for any purpose, without fee, subject to these |
| 97 |
* conditions: |
| 98 |
* (1) If any part of the source code for this software is distributed, then this |
| 99 |
* README file must be included, with this copyright and no-warranty notice |
| 100 |
* unaltered; and any additions, deletions, or changes to the original files |
| 101 |
* must be clearly indicated in accompanying documentation. |
| 102 |
* (2) If only executable code is distributed, then the accompanying |
| 103 |
* documentation must state that "this software is based in part on the work of |
| 104 |
* the Independent JPEG Group". |
| 105 |
* (3) Permission for use of this software is granted only if the user accepts |
| 106 |
* full responsibility for any undesirable consequences; the authors accept |
| 107 |
* NO LIABILITY for damages of any kind. |
| 108 |
* |
| 109 |
* These conditions apply to any software derived from or based on the IJG code, |
| 110 |
* not just to the unmodified library. If you use our work, you ought to |
| 111 |
* acknowledge us. |
| 112 |
* |
| 113 |
* Permission is NOT granted for the use of any IJG author's name or company name |
| 114 |
* in advertising or publicity relating to this software or products derived from |
| 115 |
* it. This software may be referred to only as "the Independent JPEG Group's |
| 116 |
* software". |
| 117 |
* |
| 118 |
* We specifically permit and encourage the use of this software as the basis of |
| 119 |
* commercial products, provided that all warranty or liability claims are |
| 120 |
* assumed by the product vendor. |
| 121 |
*/ |
| 122 |
|
| 123 |
Getting the JPEG compression working was not the end of the story, after this |
| 124 |
I could create .dpf files which the windows software would read. But the files |
| 125 |
had a "mystery" block in them between the 16 byte header and the quantisation |
| 126 |
tables, and the picframe does not grok the files without the mystery block |
| 127 |
containing the proper data. It turns out this mystery block contains an 8 byte |
| 128 |
info block per MCU which makes it possible to start decoding with a random |
| 129 |
MCU, rather then needing to do the decode linearly. This way the picture frame |
| 130 |
can do transition effects. The info in each MCU info block consists of |
| 131 |
last DC vals for the 3 components and the offset in the huffman bitstream |
| 132 |
where the data for this block starts. For more info so the dump of the wiki |
| 133 |
page below. |
| 134 |
|
| 135 |
|
| 136 |
Below is a dump of: |
| 137 |
http://picframe.spritesserver.nl/wiki/index.php/ImageEncodingAx206 |
| 138 |
Made on 30 March 2010. |
| 139 |
|
| 140 |
Contents |
| 141 |
|
| 142 |
• 1 Compressed image format |
| 143 |
□ 1.1 Introduction |
| 144 |
□ 1.2 Image format |
| 145 |
☆ 1.2.1 Header |
| 146 |
☆ 1.2.2 MCU info block |
| 147 |
☆ 1.2.3 DQT table |
| 148 |
☆ 1.2.4 DHT table |
| 149 |
☆ 1.2.5 Entropy coded data |
| 150 |
|
| 151 |
Compressed image format |
| 152 |
|
| 153 |
This page describes the currently known information about the compression of |
| 154 |
image files used in pictureframes using the ax206 chip. |
| 155 |
|
| 156 |
Introduction |
| 157 |
|
| 158 |
The ax206 chip stores images in a JPEG-like format, but with several |
| 159 |
modifications and/or parts missing. Images also contain some kind of thumbnail |
| 160 |
image. |
| 161 |
|
| 162 |
To understand the information on this page, it is recommended to have a basic |
| 163 |
understanding of JPEG compression. |
| 164 |
|
| 165 |
Image format |
| 166 |
|
| 167 |
The image file seems to contain the following parts: |
| 168 |
|
| 169 |
• a 16-byte header |
| 170 |
• a small, low-res version of the image in a kind of raw format (color space |
| 171 |
is unknown, but likely YCbCr) |
| 172 |
• a JPEG DQT table (but without the DQT marker) containing two JPEG |
| 173 |
quantisation tables, one for luma and one for chroma |
| 174 |
• a JPEG DHT table (but without the DHT marker) containing four huffman |
| 175 |
tables (combinations of luma/chroma, DC/AC) |
| 176 |
• the entropy coded JPEG data |
| 177 |
|
| 178 |
Although the encoding is very much like JPEG, there are no JPEG chunk markers |
| 179 |
at all. |
| 180 |
|
| 181 |
Header |
| 182 |
|
| 183 |
The image header is contained in the first 16 bytes. |
| 184 |
|
| 185 |
Some of the fields in the header can be related to values from the JPEG |
| 186 |
specification. |
| 187 |
|
| 188 |
The table below shows all known fields: |
| 189 |
|
| 190 |
┌───────┬─────────┬───────────────────────────────────────────────────────────┐ |
| 191 |
│ │ Related │ │ |
| 192 |
│ Byte │ JPEG │ Meaning │ |
| 193 |
│ │ field │ │ |
| 194 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 195 |
│0x00/ │SOF: X │Image width, encoded as big-endian │ |
| 196 |
│0x01 │ │ │ |
| 197 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 198 |
│0x02/ │SOF: Y │Image height, encoded as big-endian │ |
| 199 |
│0x03 │ │ │ |
| 200 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 201 |
│ │SOF: Hi/ │"Resolution" byte, 0 for images without sub-sampling of the│ |
| 202 |
│0x04 │Vi │color components, 3 for images that use sub-sampling of the│ |
| 203 |
│ │ │color components │ |
| 204 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 205 |
│0x05/ │ │Three numbers, each indicating which quantisation table (0 │ |
| 206 |
│0x06/ │DQT: Tq │or 1) should be used for which component (Y, Cb, Cr) │ |
| 207 |
│0x07 │ │ │ |
| 208 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 209 |
│0x08/ │ │Three numbers, each indicating which DC huffman table (0 or│ |
| 210 |
│0x09/ │SOS: Tdj │1) should be used for which component (Y, Cb, Cr) │ |
| 211 |
│0x0A │ │ │ |
| 212 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 213 |
│0x08/ │ │Three numbers, each indicating which AC huffman table (0 or│ |
| 214 |
│0x09/ │SOS: Taj │1) should be used for which component (Y, Cb, Cr) │ |
| 215 |
│0x0A │ │ │ |
| 216 |
├───────┼─────────┼───────────────────────────────────────────────────────────┤ |
| 217 |
│0x0E/ │Unknown │Unknown │ |
| 218 |
│0x0F │ │ │ |
| 219 |
└───────┴─────────┴───────────────────────────────────────────────────────────┘ |
| 220 |
|
| 221 |
MCU info block |
| 222 |
|
| 223 |
Immediately after the header there is an MCU info block, this block contains 8 |
| 224 |
bytes per MCU which makes it possible to start decoding with a random MCU, |
| 225 |
rather then needing to do the decode linearly. This way the picture frame can |
| 226 |
do transition effects. The size of this block depends on the number of MCU's |
| 227 |
and thus on the "resolution" byte 0x04 from the header. For a resolution value |
| 228 |
of 0x00, the size (in bytes) is width * height / 8. For a resolution value of |
| 229 |
0x03, the size is width * height / 32. |
| 230 |
|
| 231 |
The info in each MCU info block consists of last DC vals for the 3 components |
| 232 |
and the location in the huffman bitstream where the data for this block starts. |
| 233 |
The way the location in the huffman bitstream is coded is, erm, rather |
| 234 |
interesting. The contents of this field of the per MCU info is an offset to add |
| 235 |
to the start of the MCU info block. An example to make things more clear. Lets |
| 236 |
say the huffman compressed data for the 4th MCU starts at location 3066 |
| 237 |
(decimal), the information block for the 4th MCU starts at offset 16 + 3 * 8 = |
| 238 |
40 (decimal). So the contents of the field coding were the huffman compressed |
| 239 |
data for the 4th MCU starts is 3066 - 40 = 3026. So say that the 5th MCU starts |
| 240 |
5 bytes later (the 4th one compressed well), then the field coding were the |
| 241 |
huffman compressed data starts would contain 3071 - 48 = 3023. Yes this is a |
| 242 |
bit weird, but it is how it is. |
| 243 |
|
| 244 |
┌─────────┬───────────────────────────────────────────────────────────────────┐ |
| 245 |
│ Offset │ Meaning │ |
| 246 |
├─────────┼───────────────────────────────────────────────────────────────────┤ |
| 247 |
│0x00-0x01│Little Endian 16 bit word storing last DC val for the Y channel │ |
| 248 |
├─────────┼───────────────────────────────────────────────────────────────────┤ |
| 249 |
│0x02-0x03│Little Endian 16 bit word storing last DC val for the Cb channel │ |
| 250 |
├─────────┼───────────────────────────────────────────────────────────────────┤ |
| 251 |
│0x04-0x05│Little Endian 16 bit word storing last DC val for the Cr channel │ |
| 252 |
├─────────┼───────────────────────────────────────────────────────────────────┤ |
| 253 |
│ │Little Endian 16 bit word storing the offset into the image, from │ |
| 254 |
│0x06-0x07│the start of this info block! Where the huffman data for this MCU │ |
| 255 |
│ │starts │ |
| 256 |
└─────────┴───────────────────────────────────────────────────────────────────┘ |
| 257 |
|
| 258 |
DQT table |
| 259 |
|
| 260 |
This table is basically a standard JPEG DQT table, but it seems to be missing |
| 261 |
its 0xFFDB marker at the start. |
| 262 |
|
| 263 |
DHT table |
| 264 |
|
| 265 |
This table is basically a standard JPEG DHT table, but it seems to be missing |
| 266 |
its 0xFFC4 marker at the start. |
| 267 |
|
| 268 |
Entropy coded data |
| 269 |
|
| 270 |
This seems to be standard huffman encoded coefficient data, with the following |
| 271 |
exceptions: |
| 272 |
|
| 273 |
• there are no stuffing bytes after 0xFF bytes |
| 274 |
• after each JPEG MCU (minimum coded unit), the next MCU seems to start at |
| 275 |
the next byte-aligned position in the stream |
| 276 |
• there is no end-of-image JPEG marker |
| 277 |
|
| 278 |
The order of components (Y, Cb, Cr) coded in the stream depends on the |
| 279 |
resolution byte from the header. |
| 280 |
|
| 281 |
• For resolution value 0x00, they seem to be interleaved in the following |
| 282 |
order: chroma1, chroma2, luma. |
| 283 |
• For resolution value 0x03, they seem to be interleaved in the following |
| 284 |
order: chroma1, chroma2, luma, luma, luma, luma. The two chroma blocks are |
| 285 |
sub-sampled by a factor 2 in this case (and therefore encode for a 16x16 |
| 286 |
pixel area), the luma blocks are not sub-sampled (so that's why there are 4 |
| 287 |
blocks of 8x8 pixels to cover the same 16x16 pixel area). |
| 288 |
|
| 289 |
Retrieved from "http://picframe.spritesserver.nl/wiki/index.php/ |
| 290 |
ImageEncodingAx206" |
| 291 |
|
| 292 |
Powered by MediaWiki |
| 293 |
GNU Free Documentation License 1.2 |
| 294 |
|
| 295 |
• This page was last modified 14:09, 30 March 2010. |
| 296 |
• This page has been accessed 29 times. |
| 297 |
• Content is available under GNU Free Documentation License 1.2. |