New lossless compressor for 24-bit images (3 channels, 8 bits per channel)

**Alexander Rhatushnyak** · 21st July 2010, 06:46

Demo version is here:
http://www.imagecompression.info/gralic/Gralic1d.zip
It can handle only images in PNM or PPM format, 24-bit or 8-bit, with no comments,
with both width and height less than 8184.

It compresses some image sets better and faster than BMF 2, paq8px, and others.

For example, all 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175652563 -- GraLIC 1.3.demo

Another example, all 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
107479066 -- GraLIC 1.3.demo

GraLIC 1.3 compresses images better (on average) than 1.3.demo, for example this set - http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D - is 59741781 bytes compressed by GraLIC 1.3 :

http://www.imagecompression.info/gra...f_paradise.bin
http://www.imagecompression.info/gra..._cathedral.bin
http://www.imagecompression.info/gra...nolulu_zoo.bin
http://www.imagecompression.info/gra...northcoast.bin
http://www.imagecompression.info/gra...ahu_waimea.bin
http://www.imagecompression.info/gra...hu_waimea2.bin
http://www.imagecompression.info/gra...i_at_night.bin

GraLIC 1.3.demo is able to decompress these files, but it compresses this set to 61483334 bytes (2.9% bigger).

**ddfox** · 22nd July 2010, 10:35

Thanks. downloading testfile......

**Matt Mahoney** · 22nd July 2010, 16:48

Some results with zpaq ocbmp_j4.cfg on the imagecompression.info 8 bit color benchmark converted to .bmp. Compression speed is about 9-10 seconds per MB on a 2 GHz T3200/Vista 32 bit

Code:

    609,836 artificial.bmp.zpaq
 46,563,935 big_building.bmp.zpaq
 39,867,729 big_tree.bmp.zpaq
 17,854,911 bridge.bmp.zpaq
  7,032,186 cathedral.bmp.zpaq
 19,273,915 deer.bmp.zpaq
  3,398,567 fireworks.bmp.zpaq
  1,801,623 flower_foveon.bmp.zpaq
  5,018,225 hdr.bmp.zpaq
  8,732,160 leaves_iso_1600.bmp.zpaq
  6,861,988 leaves_iso_200.bmp.zpaq
  4,919,898 nightshot_iso_100.bmp.zpaq
  9,750,463 nightshot_iso_1600.bmp.zpaq
  6,314,328 spider_web.bmp.zpaq
 177,999,764 bytes

**Alexander Rhatushnyak** · 22nd July 2010, 18:42

Originally Posted by Matt Mahoney

Compression speed is about 9-10 seconds per MB on a 2 GHz T3200/Vista 32 bit

How does it compare with GraLIC's speed?
Here's what I got compressing the smallest file, flower_foveon:

C:\MinGW\bin>zpaq ocbmp_j4.cfg qfqfqf FLOWER~1.BMP
107.932 MB memory required.
C:\DOCUME~1\ADMINI~1\LOCALS~1\Temp\zpaq_ec6c24c399 07fb335ad4d3f9e89073603b1cda60.exe ocbmp_j4.cfg qfqfqf FLOWER~1.BMP
zpaq rcolorpre.cfg FLOWER~1.BMP C:\DOCUME~1\ADMINI~1\LOCALS~1\Temp\qfqfqf.zpaq.pre ... Process time 2.47 sec. Wall time 3 sec.
OK
FLOWER~1.BMP 10287702 -> 1801603
-> 1801603

Memory utilization:
0 const 160
1 cm 20 255: 52403/1048576 (5.00%)
2 cm 20 255: 61810/1048576 (5.89%)
3 cm 20 255: 68229/1048576 (6.51%)
4 cm 20 255: 66507/1048576 (6.34%)
5 cm 20 255: 94109/1048576 (8.97%)
6 cm 20 255: 87452/1048576 (8.34%)
7 cm 20 255: 72799/1048576 (6.94%)
8 cm 20 255: 65998/1048576 (6.29%)
9 icm 16: 67461/4194304 (1.61%)
10 cm 20 255: 61397/1048576 (5.86%)
11 icm 16: 65299/4194304 (1.56%)
12 cm 20 255: 535839/1048576 (51.10%)
13 icm 16: 1948244/4194304 (46.45%)
14 cm 20 255: 529438/1048576 (50.49%)
15 icm 16: 296856/4194304 (7.08%)
16 cm 20 255: 233900/1048576 (22.31%)
17 cm 20 255: 1048096/1048576 (99.95%)
18 cm 20 255: 56143/1048576 (5.35%)
19 cm 20 255: 1048220/1048576 (99.97%)
20 cm 20 255: 58111/1048576 (5.54%)
21 icm 16: 1591761/4194304 (37.95%)
22 icm 16: 1590658/4194304 (37.92%)
23 cm 11 255: 757/2048 (36.96%)
24 mix 16 0 24 16 255: 183673/1572864 (11.68%)
25 mix 11 0 25 20 255: 18763/51200 (36.65%)
26 mix2 0 24 25 40 0: 1/1 (100.00%)
27 mix 0 0 26 32 0: 26/26 (100.00%)
28 mix2 0 26 27 40 0: 1/1 (100.00%)
29 sse 16 28 8 255: 104004/2097152 (4.96%)
30 mix2 8 28 29 40 255: 255/256 (99.61%)
31 sse 8 30 8 255: 5423/8192 (66.20%)
32 mix2 0 30 31 40 0: 1/1 (100.00%)
Process time 45.55 sec. Wall time 46 sec.
Process time 45.56 sec. Wall time 46 sec.

C:\MinGW\bin>GraLIC1d.exe c qwqwq2 flower_foveon.ppm
Encoding flower_foveon.ppm 10287665->1767310, 1.3743 bpc, 4.39 sec. 2343 kb/s

~10.5 times slower. Different for bigger files and/or on another computer?

BTW, GraLIC used less than 30 Mb of memory; no temporary files.

**Matt Mahoney** · 22nd July 2010, 23:01

On my computer, GraLIC takes 7.2 sec to compress, 8.3 to decompress. zpaq ocbmp_j4.cfg takes 82.4 sec, 79.8 sec.

**m^2** · 23rd July 2010, 10:55

When can we expect a full version?
Do you have plans to drop the size / colour limitations?

**m^2** · 24th July 2010, 00:36

I made a quick and dirty test of files that are somewhat representative to what I usually work with, high resolution scans of printed media. Dots can't be seen individually, but make the images somewhat noisy.
22 files.
BMP:
440327572 B
PNGOUT Sep 22 2009 /s0:
180381473 B
BMF 2.0 -s:
114553144 B
GraLIC1d:
123665770 B

I didn't measure time, but GraLIC was clearly the fastest (and pngout the slowest), but on this image set it was far from being stronger than BMF.

**Alexander Rhatushnyak** · 24th July 2010, 01:19

Originally Posted by m^2

When can we expect a full version?

Hopefully soon. What is the minimal set of features a full version must have?

Originally Posted by m^2

22 files.

Are all of them 24-bit RGB images? Are there any decompressed JPGs, GIFs, etc. among them? Can you please send the image with the biggest GraLIC_size/BMF_size ratio?

**m^2** · 24th July 2010, 12:09

Originally Posted by Alexander Rhatushnyak

Hopefully soon. What is the minimal set of features a full version must have?

Only stability and established file format.

Originally Posted by Alexander Rhatushnyak

Are all of them 24-bit RGB images? Are there any decompressed JPGs, GIFs, etc. among them? Can you please send the image with the biggest GraLIC_size/BMF_size ratio?

All are 24 bit scans with little processing, no lossy compression. I deleted the gralic / bmf files so I have to compress them again to find out, I'll PM you the file when it's done.

**Alexander Rhatushnyak** · 24th July 2010, 19:12

Originally Posted by m^2

Only stability and established file format.

Both depend on how much the current release is tested, how many problems are reported.

Originally Posted by m^2

All are 24 bit scans with little processing

Processing like noise reduction, smoothing or sharpening?
Such scans look neither photographic nor artificial to GraLIC 1.3.d, I guess, that's why it makes the wrong choice so often.

Originally Posted by m^2

I'll PM you the file when it's done.

Thank you!
Actually, I believe the more such images I get, the better GraLIC will compress such images

**m^2** · 25th July 2010, 01:57

Originally Posted by Alexander Rhatushnyak

Both depend on how much the current release is tested, how many problems are reported.

OK.

Originally Posted by Alexander Rhatushnyak

Processing like noise reduction, smoothing or sharpening?
Such scans look neither photographic nor artificial to GraLIC 1.3.d, I guess, that's why it makes the wrong choice so often.
Thank you!

Varies. Sometimes the dots are smoothed somewhat, but usually they aren't. Actually I can never tell what exactly was done because I'm not the one who scans them. But I'm positive that good majority has nothing else than downsizing (rarely), rotating, stuff like contrast.

Originally Posted by Alexander Rhatushnyak

Actually, I believe the more such images I get, the better GraLIC will compress such images

OK, I'll send you a bigger pack tomorrow.

**m^2** · 25th July 2010, 13:09

1. Bug: When trying to compress image that's too large, GraLIC crashes instead of showing error.
2. It would be better not to write a file in case of error.
3. What is "Unsupported PNM header."?

**Alexander Rhatushnyak** · 30th July 2010, 23:54

Version 1.4.demo was released today:
http://www.imagecompression.info/gralic/Gralic1d.zip
Does not crash if width or height are bigger than 8160;
compression quality was improved.

All 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175652563 -- GraLIC 1.3.demo
175367159 -- GraLIC 1.4.demo

All 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
107479066 -- GraLIC 1.3.demo
106607166 -- GraLIC 1.4.demo

All 7 images from http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D :
78414698 -- paq8px_v69 -8 *.ppm
61332105 -- paq8px_v69 -8 *.bmp
59978252 -- BMF 2.01 -q9 -s
59967488 -- BMF 2.01 -q1 -s
59741781 -- GraLIC 1.3
59673365 -- GraLIC 1.4.demo
58619786 -- GraLIC 1.4 -- http://www.imagecompression.info/gralic/Honolulu.zip

Obsolete version 1.3.demo, in case you need it to decompress images,
is now here: http://www.imagecompression.info/gralic/Gralic13d.zip

**encode** · 31st July 2010, 00:28

c:\Test>gralic c bliss.x bliss.bmp
Encoding bliss.bmp Unsupported PNM header.
Failed to compress file.
c:\Test>gralic c Longhorn.x Longhorn.bmp
Encoding Longhorn.bmp Unsupported PNM header.
Failed to compress file.
c:\Test>gralic c rafale.x rafale.bmp
Encoding rafale.bmp Unsupported PNM header.
Failed to compress file.
c:\Test>gralic c rafale.pnm rafale.bmp
Encoding rafale.bmp Unsupported PNM header.
Failed to compress file.
c:\Test>

**Alexander Rhatushnyak** · 31st July 2010, 00:59

Originally Posted by encode

Encoding rafale.bmp

If you run it with no commands and no options, it prints

GraLIC lossless image compressor, ver. 1.4.demo
(C) 2009-2010, Alexander Rhatushnyak, Graystone Compression Technologies Inc.
This executable is for 8-bit or 24-bit images with both width and height <=8160
Please feel free to email pqr(alpha)rogers.com and bfrusina(alpha)rogers.com for more info.

To compress|decompress a PNM file: GraLIC14d.exe c|d archive file

As you can see in the first post,
GraLIC can handle only images in PNM or PPM format, 24-bit or 8-bit, with no comments, with both width and height less than 8184.
( less than 8160 for versions 1.4+.demo )

Originally Posted by m^2

2. It would be better not to write a file in case of error.
3. What is "Unsupported PNM header."?

2. sometimes, not always
3. either there's a comment between "P6" and "width height", or no "<EOL>255<EOL>" at the end of the header, or header size plus image size is not equal to the PNM file size.

Three samples of PNMs with correct headers are in the attached archive.

**m^2** · 31st July 2010, 17:41

I think it would be useful if Gralic (maybe in a verbose mode) told what's really the problem, the only thing 'unsupported' tells is that it's probably not a bug, but doesn't help solve the issue.

And what's the difference between gralic 1.4 and gralic 1.4 demo?

**Alexander Rhatushnyak** · 4th August 2010, 05:45

Originally Posted by m^2

And what's the difference between gralic 1.4 and gralic 1.4 demo?

1.4 compresses images better than 1.4.demo, on average (you are welcome to decompress and then compress images from http://www.imagecompression.info/gralic/Honolulu.zip with GraLIC 1.4.demo)
1.4 can compress images with width and height up to 65500,
1.4 can be customized for much lower memory usage (const1+max(width,height)*const2) for much higher speed (once the target platform is specified), for certain types of images, etc.

**m^2** · 4th August 2010, 17:21

So can we get 1.4?

**Alexander Rhatushnyak** · 10th August 2010, 04:31

Version 1.5.demo is here:
http://www.imagecompression.info/gralic/Gralic1d.zip
Better support for PNMs with comments (see the header in Gralic1d.zip\example.gra15 - decompress and open it in text editor).
Unlike version 1.4.d, new version compresses some artificial images much better than BMF and PAQ8,
for example http://cdb.paradice-insight.us/corpo...falten.ppm.bz2
On photographic images, compression quality got better, on average:

All 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175367159 -- GraLIC 1.4.demo
175286083 -- GraLIC 1.5.demo

All 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
106607166 -- GraLIC 1.4.demo
106111127 -- GraLIC 1.5.demo

All 7 images from http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D :
78414698 -- paq8px_v69 -8 *.ppm
61332105 -- paq8px_v69 -8 *.bmp
59978252 -- BMF 2.01 -q9 -s
59967488 -- BMF 2.01 -q1 -s
59673365 -- GraLIC 1.4.demo
59535532 -- GraLIC 1.5.demo
58617827 -- GraLIC 1.5 -- http://www.imagecompression.info/gralic/Honolulu.zip

Obsolete packages:
http://www.imagecompression.info/gralic/Gralic13d.zip
http://www.imagecompression.info/gralic/Gralic14d.zip
http://www.imagecompression.info/gra...lulu_gra14.zip

Originally Posted by m^2

So can we get 1.4?

Please e-mail to pqr(alpha)rogers-kom and bfrusina(alpha)rogers-kom for more information on GraLIC.

**Ethatron** · 13th August 2010, 13:59

Other results here:
http://cdb.paradice-insight.us/?stats=corp

**Alexander Rhatushnyak** · 23rd August 2010, 07:54

Version 1.6.demo is here:
http://www.imagecompression.info/gralic/Gralic1d.zip
Compression quality is better, on average:

All 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175286083 -- GraLIC 1.5.demo
175219620 -- GraLIC 1.6.demo

All 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
106111127 -- GraLIC 1.5.demo
106048217 -- GraLIC 1.6.demo

All 7 images from http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D :
78414698 -- paq8px_v69 -8 *.ppm
61332105 -- paq8px_v69 -8 *.bmp
59978252 -- BMF 2.01 -q9 -s
59967488 -- BMF 2.01 -q1 -s
59535532 -- GraLIC 1.5.demo
58705316 -- GraLIC 1.6.demo
58451644 -- GraLIC 1.6 -- http://www.imagecompression.info/gralic/Honolulu.zip

Obsolete packages:
http://www.imagecompression.info/gralic/Gralic13d.zip
http://www.imagecompression.info/gralic/Gralic14d.zip
http://www.imagecompression.info/gralic/Gralic15d.zip

**Alexander Rhatushnyak** · 1st September 2010, 07:30

Version 1.7.demo is here:
http://www.imagecompression.info/gralic/Gralic1d.zip
Compression quality got better, on average:

All 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175219620 -- GraLIC 1.6.demo
175046402 -- GraLIC 1.7.demo

All 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
106048217 -- GraLIC 1.6.demo
105935305 -- GraLIC 1.7.demo

All 7 images from http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D :
78414698 -- paq8px_v69 -8 *.ppm
61332105 -- paq8px_v69 -8 *.bmp
59978252 -- BMF 2.01 -q9 -s
59967488 -- BMF 2.01 -q1 -s
58705316 -- GraLIC 1.6.demo
58396278 -- GraLIC 1.7.demo

All 10 images from http://cdb.paradice-insight.us/corpo...tural/?C=S;O=D :
111728355 -- GraLIC 1.6.demo
111601659 -- GraLIC 1.7.demo
111573380 -- BMF 2.01 -q1 -s
110878448 -- BMF 2.01 -q9 -s
110663248 -- paq8px_v69 -8
110536816 -- GraLIC 1.7 - http://www.imagecompression.info/gralic/Natural8.zip -- GraLIC 1.7.demo can decompress these images.

Obsolete packages:
http://www.imagecompression.info/gralic/Gralic13d.zip
http://www.imagecompression.info/gralic/Gralic14d.zip
http://www.imagecompression.info/gralic/Gralic15d.zip
http://www.imagecompression.info/gralic/Gralic16d.zip

**Alexander Rhatushnyak** · 1st September 2010, 21:09

Originally Posted by m^2

All are 24 bit scans with little processing, no lossy compression.

Sorry, you are wrong, CD3 in your set is definitely an ex-JPEG. Can you see 8x8 squares in the attached image?

**m^2** · 1st September 2010, 21:14

Originally Posted by Alexander Rhatushnyak

Sorry, you are wrong, CD3 in your set is definitely an ex-JPEG. Can you see 8x8 squares in the attached image?

Indeed, thanks for the info.

**Alexander Rhatushnyak** · 21st September 2010, 23:41

Version 1.8.demo is here:
http://www.imagecompression.info/gralic/Gralic1d.zip
Compression quality is better, on average:

All 14 images from http://www.imagecompression.info/tes...es/rgb8bit.zip :
193508124 -- BMF 2.01 -q1
193254688 -- BMF 2.01 -q9
179124342 -- paq8px_v69 -8
176768000 -- BMF 2.01 -s -q1
176417720 -- BMF 2.01 -s -q9
175046402 -- GraLIC 1.7.demo
174768218 -- GraLIC 1.8.demo

All 11 images from http://cdb.paradice-insight.us/corpo...verse/?C=S;O=D :
109579077 -- paq8px_v69 -8
108038840 -- BMF 2.01 -q1 -s
107958888 -- BMF 2.01 -q9 -s
105935305 -- GraLIC 1.7.demo
105701138 -- GraLIC 1.8.demo

All 7 images from http://cdb.paradice-insight.us/corpo...08bit/?C=S;O=D :
78414698 -- paq8px_v69 -8 *.ppm
61332105 -- paq8px_v69 -8 *.bmp
59978252 -- BMF 2.01 -q9 -s
59967488 -- BMF 2.01 -q1 -s
58396278 -- GraLIC 1.7.demo
57942094 -- GraLIC 1.8.demo

All 10 images from http://cdb.paradice-insight.us/corpo...tural/?C=S;O=D :
111601659 -- GraLIC 1.7.demo
111573380 -- BMF 2.01 -q1 -s
111350703 -- GraLIC 1.8.demo
110878448 -- BMF 2.01 -q9 -s
110663248 -- paq8px_v69 -8
110324898 -- GraLIC 1.8 - http://www.imagecompression.info/gralic/Natural8.zip -- GraLIC 1.8.demo can decompress these images.

Obsolete packages:
http://www.imagecompression.info/gralic/Gralic13d.zip
http://www.imagecompression.info/gralic/Gralic14d.zip
http://www.imagecompression.info/gralic/Gralic15d.zip
http://www.imagecompression.info/gralic/Gralic16d.zip
http://www.imagecompression.info/gralic/Gralic17d.zip

**Matt Mahoney** · 22nd September 2010, 01:29

Some results with included example.gra18

Code:

   18,605 example.gra18
6,912,017 example.pmn
6,912,054 example.bmp (converted with ImageMagick, 1920 x 1200 x 24 bit color)
   65,255 example.png
   16,860 example.bmp.paq8px_v69-6
   15,919 example.bmp.zpaq-bmp_j4.cfg

**Alexander Rhatushnyak** · 22nd September 2010, 01:50

Compression and decompression times.

On a 7 Mb image (London.ppm from the 4th set, 6888017 bytes)
66+67 seconds -- paq8px_v69 -8
12+~11 sec -- bmf -s
27+11 sec -- bmf -s -q9
4+4 sec -- GraLIC 1.8.demo

On a 18 Mb image (Cathedral.ppm from the 1st set, 18048017 bytes)
183+185 -- paq8px_v69 -8
29+~25 -- bmf -s
62+25 -- bmf -s -q9
12+13 -- GraLIC 1.8.demo

On a 83 Mb image (Big_tree.ppm from the 1st set, 83101217 bytes)
919+928 -- paq8px_v69 -8
110+~89 -- bmf -s
425+89 -- bmf -s -q9
63+70 -- GraLIC 1.8.demo

~ means approximately

Raw output and scripts used are here, if you wish: http://www.imagecompression.info/gralic/cd_times.zip

**Matt Mahoney** · 22nd September 2010, 02:53

I did some tests on the Kodak set. For gralic I converted bmp -> ppm using ImageMagick. Timed on a 2 GHz T3200, 3 GB, 32 bit Vista.

Code:

ImageMagick convert -> png
 779,954 kodim01.bmp.png (0.5 sec each)
 618,548 kodim02.bmp.png
 544,790 kodim03.bmp.png
 638,269 kodim04.bmp.png
 807,247 kodim05.bmp.png
 669,846 kodim06.bmp.png
 569,418 kodim07.bmp.png
 787,948 kodim08.bmp.png
 584,803 kodim09.bmp.png
 594,892 kodim10.bmp.png
 638,798 kodim11.bmp.png
 571,267 kodim12.bmp.png
 889,850 kodim13.bmp.png
 739,493 kodim14.bmp.png
 614,379 kodim15.bmp.png
 576,726 kodim16.bmp.png
 625,312 kodim17.bmp.png
 779,719 kodim18.bmp.png
 671,361 kodim19.bmp.png
 508,595 kodim20.bmp.png
 680,320 kodim21.bmp.png
 701,178 kodim22.bmp.png
 559,061 kodim23.bmp.png
 707,162 kodim24.bmp.png
15,858,936 bytes

gralic18d
 467,155 kodim01.bmp.ppm.gra18 (1 sec each)
 413,996 kodim02.bmp.ppm.gra18
 328,054 kodim03.bmp.ppm.gra18
 407,877 kodim04.bmp.ppm.gra18
 468,590 kodim05.bmp.ppm.gra18
 410,285 kodim06.bmp.ppm.gra18
 336,034 kodim07.bmp.ppm.gra18
 509,919 kodim08.bmp.ppm.gra18
 362,855 kodim09.bmp.ppm.gra18
 371,716 kodim10.bmp.ppm.gra18
 389,214 kodim11.bmp.ppm.gra18
 367,652 kodim12.bmp.ppm.gra18
 532,655 kodim13.bmp.ppm.gra18
 470,170 kodim14.bmp.ppm.gra18
 378,510 kodim15.bmp.ppm.gra18
 359,454 kodim16.bmp.ppm.gra18
 369,479 kodim17.bmp.ppm.gra18
 490,024 kodim18.bmp.ppm.gra18
 447,387 kodim19.bmp.ppm.gra18
 310,953 kodim20.bmp.ppm.gra18
 426,102 kodim21.bmp.ppm.gra18
 448,625 kodim22.bmp.ppm.gra18
 351,454 kodim23.bmp.ppm.gra18
 420,405 kodim24.bmp.ppm.gra18
9,838,565 bytes

paq8px_v69 -6
 336,531 kodim01.bmp.paq8px (20 sec each)
 272,469 kodim02.bmp.paq8px
 228,596 kodim03.bmp.paq8px
 289,200 kodim04.bmp.paq8px
 391,600 kodim05.bmp.paq8px
 312,026 kodim06.bmp.paq8px
 254,876 kodim07.bmp.paq8px
 391,435 kodim08.bmp.paq8px
 268,873 kodim09.bmp.paq8px
 276,383 kodim10.bmp.paq8px
 302,321 kodim11.bmp.paq8px
 251,392 kodim12.bmp.paq8px
 430,983 kodim13.bmp.paq8px
 350,764 kodim14.bmp.paq8px
 281,368 kodim15.bmp.paq8px
 256,910 kodim16.bmp.paq8px
 276,108 kodim17.bmp.paq8px
 404,114 kodim18.bmp.paq8px
 318,057 kodim19.bmp.paq8px
 257,988 kodim20.bmp.paq8px
 321,707 kodim21.bmp.paq8px
 365,020 kodim22.bmp.paq8px
 282,442 kodim23.bmp.paq8px
 331,970 kodim24.bmp.paq8px
7,453,133 bytes

zpaq ocbmp_j4.cfg
 338,151 kodim01.bmp.zpaq (10-11 sec each)
 272,029 kodim02.bmp.zpaq
 230,431 kodim03.bmp.zpaq
 291,064 kodim04.bmp.zpaq
 393,110 kodim05.bmp.zpaq
 314,011 kodim06.bmp.zpaq
 255,350 kodim07.bmp.zpaq
 390,582 kodim08.bmp.zpaq
 270,386 kodim09.bmp.zpaq
 278,175 kodim10.bmp.zpaq
 303,601 kodim11.bmp.zpaq
 253,364 kodim12.bmp.zpaq
 431,880 kodim13.bmp.zpaq
 353,718 kodim14.bmp.zpaq
 281,349 kodim15.bmp.zpaq
 258,687 kodim16.bmp.zpaq
 277,247 kodim17.bmp.zpaq
 404,235 kodim18.bmp.zpaq
 319,912 kodim19.bmp.zpaq
 259,391 kodim20.bmp.zpaq
 323,942 kodim21.bmp.zpaq
 365,658 kodim22.bmp.zpaq
 285,507 kodim23.bmp.zpaq
 333,469 kodim24.bmp.zpaq
7,485,249 bytes

**Alexander Rhatushnyak** · 23rd September 2010, 01:43

Originally Posted by Matt Mahoney

I did some tests on the Kodak set.

As we can see on pages
http://cdb.paradice-insight.us/?stats=corp&preset=15 and
http://cdb.paradice-insight.us/?stats=clss&preset=6 ,
PAQ8 compresses much better than GraLIC ex-PCDs and artificial images
(also ex-JPEGs though there's no such set yet)
while GraLIC is only 2% better than PAQ8 on photograpic images and on 8-bit grayscale:
http://cdb.paradice-insight.us/?stats=corp&preset=17
http://cdb.paradice-insight.us/?stats=clss&preset=1
I believe ZPAQ shows similar performance.

Thread: New lossless compressor for 24-bit images (3 channels, 8 bits per channel)

Thread Tools

Search Thread

Display

New lossless compressor for 24-bit images (3 channels, 8 bits per channel)

Similar Threads

GraLIC - new lossless image compressor

a little bit off-topic, but an useful open-source - program to mount cd/dvd - images

Encode's IRC channel details?

Creating our IRC channel

RINGS Fast Bit Compressor.

Tags for this Thread

Posting Permissions