If you specify QUALITY_OUT_MAX less than QUALITY_OUT_MIN+2, imgmin will bug out with an error message such as the following:

Before quality:94 colors:93196 size:247.6kB type:TrueColor format:JPEG
imgmin: wand/magick-wand.c:181: DestroyMagickWand: Assertion `wand !=(MagickWand *) ((void *)0)' failed.
Aborted (core dumped)

It would be better if imgmin either give a decent error message, ie. referring the user to check the out min/max numbers, or if only the min or max value is given, it could automatically push the other value to be a value which makes sense, ie. 10 apart from the value specified.

instead of focusing on a single stat, build hundreds of before/after images which can be used for testing and also a corpus for a more flexible machine learning approach.

It would be helpful to install imgmin as
apt-get install -y imgmin

Hello,

Wanted to try the project, but this won't compile.

Is it possible that in the near future you will use mozjpeg instead of imagemagick?

currently we can demonstrate imgmin/httpd automation via the mod_ext_filter, but this is too slow for production.

port imgmin to native apache filter module for performance

integrate imgmin with nginx

I needed a logo for my presentation, so I've made one for you;

need this for performance, see mod_ext_filter_imgmin.pl for existing implementation

integrate imgmin into mod_pagespeed arch. make it optional.

this will make imgmin instantly available to a wide audience.

Line 337 of imgmin.c contains:

        (void) MagickSetImageProperty(mw, "jpeg:sampling-factor", "2x2");

I find that the quality is noticeably and significantly better if I comment/remove that line (although the size is increased some).

Here are some test images.

graphic made in GIMP, saved as quality 100 JPEG:
original
imgmin output, chroma downsampling 2x2
imgmin output, chroma downsampling off

random image from rawsamples.ch, resized, saved as quality 100 JPEG:
original
imgmin output, chroma downsampling 2x2
imgmin output, chroma downsampling off

random image from Google Images (and ultimately from thewallgans.com, a wallpaper site), resized, saved as quality 100 JPEG:
original
imgmin output, chroma downsampling 2x2
imgmin output, chroma downsampling off

Every area with fine detail is blurred significantly, and colors tend to look less vivid, in the chroma-downsampled versions.

Hi. This looks like an amazing tool. And thanks for the very informative readme.

But I'm having trouble building on Mac OS X Lion...

MBP:imgmin callum$ make
make -C src
make -C apache2
apxs2 -I `pwd`/.. `MagickWand-config --cppflags` `MagickWand-config --ldflags --libs|xargs` -Wc,-DIMGMIN_LIB -Wc,-W -Wc,-Wall -Wc,-Wno-unused-parameter -c mod_imgmin.c -c ../imgmin.c
/bin/sh: apxs2: command not found
make[2]: *** [mod_imgmin.la] Error 127
make[1]: *** [mod_imgmin] Error 2
make: *** [all] Error 2

OS X does seem to have apxs already:

MBP:imgmin callum$ apxs
Usage: apxs -g [-S <var>=<val>] -n <modname>
       apxs -q [-S <var>=<val>] <query> ...
       apxs -c [-S <var>=<val>] [-o <dsofile>] [-D <name>[=<value>]]
               [-I <incdir>] [-L <libdir>] [-l <libname>] [-Wc,<flags>]
               [-Wl,<flags>] [-p] <files> ...
       apxs -i [-S <var>=<val>] [-a] [-A] [-n <modname>] <dsofile> ...
       apxs -e [-S <var>=<val>] [-a] [-A] [-n <modname>] <dsofile> ...

...but not apxs2.

Googling "mac apxs2" etc. hasn't been very helpful. Any ideas?

Thanks
Callum

allow user to set a deadline, if the deadline is reached the best image found at that point will be the result.

imgmin v1.1 When performing overflow happened, terminated by signal SIGSEGV (Address boundary error).

imgmin /usr/local/Library/Homebrew/test/fixtures/test.jpg out.jpg
Before quality:95 colors:1 size:  0.3kB type:Palette format:JPEG
terminated by signal SIGSEGV (Address boundary error)

# file '/usr/local/Library/Homebrew/test/fixtures/test.jpg'
  /usr/local/Library/Homebrew/test/fixtures/test.jpg: JPEG image data, JFIF standard 1.01

• Image overflow: https://git.io/vJ11U
• Work: https://upload.wikimedia.org/wikipedia/en/a/a9/Example.jpg
• Homebrew/legacy-homebrew#39383

Good day!
Please post a compiled version for Windows.

kornelski/imgmin@pornel:0fe0e25...pornel:d0b9172

Instead of searching for quality using bisection method, which is slow, I'm trying to find exact quality needed by modelling jpeg quality vs dssim ratio as a curve, and finding the desired point on that curve.

I think I've told my math teacher that polynomials are stupid and I'm never going to use them… but…

The curve is a polynomial constructed from 3 points, and the points represent jpeg-quality to dssim relationship from the last 3 runs (default points are used for the first couple of runs).

It needs a bit of tweaking, but when it works it finds exact quality in 3-4 steps.

images with less than, say, 4096 pixels can't have 4096 colors, so optimization is skipped. does it matter?

I have an idea on optimization of JPEG Lossy

What is JPEG progressive? Actually it is a number of images where quality of each following image is higher previous the image.

We will take this image

JPEG Scan Killer project.

We receive 16 images:

How it is possible to create with the help -scans technology of optimization of JPEG lossy?

JPEG consists of blocks of 8х8 pixels. Perhaps, there is a sense to segment the image and to optimize each part separately. Later to connect parts of the image in the uniform optimized image. Principle of work of the betterjpeg program.

I wanted to learn, your opinion on the matter.

Similar to this issue from early 2015 #51 on Ubuntu 16.04, following the build instructions precisely (apt installs ImageMagick v6.8.9) building imgmin fails. Also worth noting I had to manually export MAGICK_CONFIG=/usr/lib/x86_64-linux-gnu/ImageMagick-6.8.9/bin-Q16/Magick-config or config/make would fail to find the config and would result in an error like --cflags: command not found since the MAGICK_CONFIG default value falls back to an empty string.

The error details:

root@1edbd36cf720:/root/imgmin-1.1# make
make  all-recursive
make[1]: Entering directory '/root/imgmin-1.1'
Making all in src
make[2]: Entering directory '/root/imgmin-1.1/src'
gcc -W -Wall -Os  `/usr/bin/GraphicsMagickWand-config --cflags --cppflags` -o imgmin imgmin.c dssim.c `/usr/bin/GraphicsMagickWand-config --ldflags --libs` -lm
imgmin.c:28:29: fatal error: wand/MagickWand.h: No such file or directory
compilation terminated.
dssim.c: In function ‘convert_image_row’:
dssim.c:304:55: warning: unused parameter ‘inf’ [-Wunused-parameter]
 static void convert_image_row(const dssim_info *const inf, float *const channels[], const int num_channels, const int y, const int width, void *user_data)
                                                       ^
Makefile:566: recipe for target 'imgmin' failed
make[2]: *** [imgmin] Error 1
make[2]: Leaving directory '/root/imgmin-1.1/src'
Makefile:355: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory '/root/imgmin-1.1'
Makefile:296: recipe for target 'all' failed
make: *** [all] Error 2

I'm using Docker, phusion/baseimage:0.9.19 which is ubuntu 16.04. Here's a minimal Dockerfile that will reproduce the issue:

FROM phusion/baseimage:0.9.19

# Use baseimage-docker's init system.
CMD [ /sbin/my_init ]

# upgrade
RUN apt-get update && apt-get dist-upgrade -y

# build tools (uninstall afterwards)
RUN apt-get install -y \
  build-essential \
  dh-autoreconf \
  wget

RUN apt-get install -y libpng-dev
RUN cd && wget https://github.com/pornel/pngquant/archive/2.5.0.tar.gz && \
  tar xf 2.5.0.tar.gz && \
  cd pngquant-2.5.0 && \
  make && \
  make install

RUN apt-get install -y imagemagick libgraphicsmagick1-dev libmagickwand-dev perlmagick libmagick++-6-headers libmagick++-dev
ENV MAGICK_CONFIG=/usr/bin/GraphicsMagickWand-config
RUN cd && wget https://github.com/rflynn/imgmin/archive/v1.1.tar.gz && \
  tar xf v1.1.tar.gz && \
  cd imgmin-1.1 && \
  autoreconf -fi && \
  ./configure && \
  make && \
  make install

You can build the image with docker build -t thumbor . when in a dir with the Dockerfile above.

It will fail but you can still shell in with docker run -t -i thumbor /sbin/my_init -- bash -l. After you shell in you'd probably want to rerun imgmin build step (it fails so it won't be in the image) with cd /root && wget https://github.com/rflynn/imgmin/archive/v1.1.tar.gz && \ tar xf v1.1.tar.gz && \ cd imgmin-1.1 && \ export MAGICK_CONFIG=/usr/bin/GraphicsMagickWand-config && autoreconf -fi && \ ./configure && \ make && \ make install

Would appreciate any help. Thanks.

Hi,

I'm encountering a "make" problem on ubuntu server 13.04 (GNU/Linux 3.8.0-33-generic x86_64).

Everything seems to be fine until the make command. I was able to install imgmin successfully on a 12.04 LTS ubuntu server, but not on this new machine. I love this script it is fast and the ratio quality/weight is just wonderful I think.

Below, the output in terminal:

root@pop-front1:~/imgmin# make
make all-recursive
make[1]: Entering directory /root/imgmin' Making all in src make[2]: Entering directory/root/imgmin/src'
gcc -W -Wall -Os MagickWand-config --cflags --cppflags -o imgmin imgmin.c MagickWand-config --ldflags --libs
imgmin.c: In function ‘do_png’:
imgmin.c:389:9: warning: initialization discards ‘const’ qualifier from pointer target type [enabled by default]
imgmin.c:399:9: warning: initialization discards ‘const’ qualifier from pointer target type [enabled by default]
imgmin.c:406:9: warning: initialization discards ‘const’ qualifier from pointer target type [enabled by default]
imgmin.c:429:9: warning: initialization discards ‘const’ qualifier from pointer target type [enabled by default]
imgmin.c:368:32: warning: unused parameter ‘mw’ [-Wunused-parameter]
imgmin.c:369:50: warning: unused parameter ‘opt’ [-Wunused-parameter]
make -C apache2
make[3]: Entering directory /root/imgmin/src/apache2' if [ "apxs2" != "" ]; then \ apxs2 -Ipwd/..MagickWand-config --cppflags --ldflags --libs|xargs-Wc,-DIMGMIN_LIB -Wc,-W -Wc,-Wall -Wc,-Wno-unused-parameter -c mod_imgmin.c ../imgmin.c; \ fi Use of uninitialized value $includedir in concatenation (.) or string at (eval 9) line 1. apxs:Error: Unknown option: f. Usage: apxs -g [-S <var>=<val>] -n <modname> apxs -q [-S <var>=<val>] <query> ... apxs -c [-S <var>=<val>] [-o <dsofile>] [-D <name>[=<value>]] [-I <incdir>] [-L <libdir>] [-l <libname>] [-Wc,<flags>] [-Wl,<flags>] [-p] <files> ... apxs -i [-S <var>=<val>] [-a] [-A] [-n <modname>] <dsofile> ... apxs -e [-S <var>=<val>] [-a] [-A] [-n <modname>] <dsofile> ... make[3]: *** [mod_imgmin_la] Error 1 make[3]: Leaving directory/root/imgmin/src/apache2'
make[2]: *** [mod_imgmin] Error 2
make[2]: Leaving directory /root/imgmin/src' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory/root/imgmin'
make: *** [all] Error 2

Do you know what this could be ?

Thanks

Hello,

I was not able to compile on my system (Fedora 17), as it was failing on the call to apxs, with an error stating:

apxs:Error: Unknown option: f.

After looking at what was going on, the results on my system of MagickWand-config --cppflags --ldflags --libs|xargs resulted in the following:

-fopenmp -I/usr/include/ImageMagick -lMagickWand -lMagickCore -lMagickWand -lMagickCore

I was able to get past this by updating that portion of the Makefile with the following to strip out that option (believe real change would have to go into Makefile.am):

MagickWand-config --cppflags --ldflags --libs |xargs |sed "s/-fopenmp\s//"

Afterwards, I was able to get it to compile. I do not know if this negatively affects what is built, I have no intention of using the apache module. It might also be nice to be able to override whether the module is built by including an --disable option for apache (sorry if it is there and I did not see it, am terrible with autotools).

Thank you for this tool, am excited to see how much it will shave off my images!

-Mike

When savings is > 0 but is very small, just use the original. Often the reduction in quality is disproportional to the meager reduction in size and is not worth it.

I don't know how hard this'd be or whether or not imagemagick allows you to optimize the huffman trees/make the images progressive but I do know that jpegtran allows that to happen on its own. Adding those options would also greatly help reduce the total space required.

Crap I can't edit it.

So here's the rest of it.

The optimization passes should add that at the final step once you've found the best quality level since it definitely helps reduce the total size.

OK, it's not a ton with the example images(in the after folder). It goes down by ~4%. Which isn't insane, but still the whole idea of this thing to is to make the images as small as possible without visual quality loss.

Is it possible to convert to progressive JPEGs?

I think convert -interlace JPEG after-imgmin.jpg output.jpg will degrade quality.

I'm trying to compile on OS X 10.9 that only has some half-baked Apache included by Apple with the OS:

/Applications/Xcode.app/Contents/Developer/usr/bin/make -C apache2
if [ "apxs" != "" ]; then \
        apxs -I `pwd`/.. `MagickWand-config --cppflags --ldflags --libs|xargs` -Wc,-DIMGMIN_LIB -Wc,-W -Wc,-Wall -Wc,-Wno-unused-parameter -c mod_imgmin.c ../imgmin.c; \
    fi
/usr/share/apr-1/build-1/libtool --silent --mode=compile /Applications/Xcode.app/Contents/Developer/Toolchains/OSX10.9.xctoolchain/usr/bin/cc    -DDARWIN -DSIGPROCMASK_SETS_THREAD_MASK -I/usr/local/include -I/usr/include/apache2  -I/usr/include/apr-1   -I/usr/include/apr-1  -DIMGMIN_LIB -W -Wall -Wno-unused-parameter -I/Users/kornel/imgmin/src/apache2/.. -I/usr/local/Cellar/imagemagick/6.8.9-7/include/ImageMagick-6 -DMAGICKCORE_HDRI_ENABLE=0 -DMAGICKCORE_QUANTUM_DEPTH=16 -DMAGICKCORE_HDRI_ENABLE=0 -DMAGICKCORE_QUANTUM_DEPTH=16  -c -o mod_imgmin.lo mod_imgmin.c && touch mod_imgmin.slo
env: /Applications/Xcode.app/Contents/Developer/Toolchains/OSX10.9.xctoolchain/usr/bin/cc: No such file or directory
apxs:Error: Command failed with rc=65536
.
make[3]: *** [mod_imgmin_la] Error 1
make[2]: *** [mod_imgmin] Error 2
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

The path /Applications/Xcode.app/Contents/Developer/Toolchains/OSX10.9.xctoolchain/usr/bin/cc looks quite suspicious.

Please support imgmin --version.
I feel lost after building imgmin.

Following instructions I've hit an error on autoconf:

$ autoconf -fi
autoconf: error: invalid option `-fi'
Try `autoconf --help' for more information.

Running with separated arguments also errors:

$ autoconf -f -i
configure.ac:8: error: possibly undefined macro: AM_INIT_AUTOMAKE
      If this token and others are legitimate, please use m4_pattern_allow.
      See the Autoconf documentation.

So I've dropped -f and autoconf finishes without errors, but configure does not:

$ autoconf -i
$ ./configure 
./configure: line 2145: syntax error near unexpected token `imgmin,'
./configure: line 2145: `AM_INIT_AUTOMAKE(imgmin, 0.7)'

Hello! Nice to meed U!

I tried to build imgmin for amd64 linux but didn't succeed.
Can U, please build it for me? I know it sounds stupid, but i'm trying to do that on darwin inside docker container with a slow internet connection. It takes hours and fails.
Help me, please.

I know this project is to optimize lossy compression, but I guess there are many cases in which lossless compression actually yields better compression rate at superior quality. Examples could be backgrounds with gradients or images with unicolor areas.

Would it be possible to simply generate a PNG for such 'usual suspects' and see if its size is lower than the JPG?

We have just implemented imgmin on a project and are really happy with the over all result. But in some images we have found noticeable artefacts.

Is there some way we can decrease the threshold/tolerance when picking the quality for a given image? Just a slight tweak will hopefully do the trick.

PS. Thanks for an amazing contribution to a faster interwebz.

I've heard good things about GraphicsMagick; evaluate porting from ImageMagick to it. Measure!

http://www.graphicsmagick.org/

I've added in configure.ac

ECK_PROGS(MAGICK_CONFIG, MagickWand-config Magick-config, "/usr/lib/x86_64-linux-gnu/ImageMagick-6.8.9/bin-Q16/MagickWand-config")

then

make
make  all-recursive
make[1]: Entering directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin'
Making all in src
make[2]: Entering directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin/src'
gcc -std=gnu99 -W -Wall -Os  `/usr/lib/x86_64-linux-gnu/ImageMagick-6.8.9/bin-Q16/MagickWand-config --cflags --cppflags` -o imgmin imgmin.c dssim.c `/usr/lib/x86_64-linux-gnu/ImageMagick-6.8.9/bin-Q16/MagickWand-config --ldflags --libs` -lm
imgmin.c: In function ‘convert_row_callback’:
imgmin.c:197:51: warning: unused parameter ‘inf’ [-Wunused-parameter]
 void convert_row_callback(const dssim_info *const inf, float *const channels[], const int num_channels, const int y, const int orig_width, void *user_data) {
                                                   ^
imgmin.c:197:91: warning: unused parameter ‘num_channels’ [-Wunused-parameter]
 void convert_row_callback(const dssim_info *const inf, float *const channels[], const int num_channels, const int y, const int orig_width, void *user_data) {
                                                                                           ^
imgmin.c:197:115: warning: unused parameter ‘y’ [-Wunused-parameter]
 void convert_row_callback(const dssim_info *const inf, float *const channels[], const int num_channels, const int y, const int orig_width, void *user_data) {
                                                                                                                   ^
imgmin.c: In function ‘do_png’:
imgmin.c:429:9: warning: initialization discards ‘const’ qualifier from pointer target type
         char * const argv[] = { "pngnq", "-f", src, (char*)0};
         ^
imgmin.c:439:9: warning: initialization discards ‘const’ qualifier from pointer target type
         char * const argv[] = { "pngcrush", "-force", src, out, (char*)0 };
         ^
imgmin.c:446:9: warning: initialization discards ‘const’ qualifier from pointer target type
         char * const argv[] = { "pngquant", "-force", "256", src, (char*)0 };
         ^
imgmin.c:469:9: warning: initialization discards ‘const’ qualifier from pointer target type
         char * const argv[] = { "cp", files[0].path, dst, NULL };
         ^
imgmin.c:408:32: warning: unused parameter ‘mw’ [-Wunused-parameter]
 static void do_png(MagickWand *mw, const char *src, const char *dst,
                                ^
imgmin.c:409:50: warning: unused parameter ‘opt’ [-Wunused-parameter]
                     const struct imgmin_options *opt)
                                                  ^
dssim.c: In function ‘convert_image_row’:
dssim.c:304:55: warning: unused parameter ‘inf’ [-Wunused-parameter]
 static void convert_image_row(const dssim_info *const inf, float *const channels[], const int num_channels, const int y, const int width, void *user_data)
                                                       ^
make[2]: Leaving directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin/src'
make[2]: Entering directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin'
make[2]: Leaving directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin'
make[1]: Leaving directory '/root/debian-server-tools/monitoring/cpu-speed/x/imgmin'

It produced imgmin executable.

figure out some combination of build parameters against imagemagick on osx

Would imgmin benefit from - losslessly - pack-ing JPEG-s before or after?
http://packjpg.encode.ru/?page_id=17
code: https://github.com/szepeviktor/packJPG

also, detect whether to use apxs or apxs2.

i'll need autoconf and friends to build a custom Makefile. not my strong suit.

after installing packages
sudo apt-get install imagemagick libgraphicsmagick1-dev perlmagick apache2-prefork-dev

I get magickwand errors on stock ubuntu machines

No package 'MagickWand' found
In file included from imgmin.c:26:0:
imgmin.h:7:29: fatal error: wand/MagickWand.h: No such file or directory
compilation terminated.

I've made my SSIM implementaiton usable directly in C (ignore the CLI tool that reads PNG, just set bitmaps directly via C API).

This is much more accurate than ImageMagick's MSE metric.

It may even be faster, because my implementation doesn't need to re-decode source image (you can set it once and only swap modified version).

I can't find my really old version of imgmin to compare but I just ran an image through it and it set the after quality to 65 as the "perfect" quality even though the image is full of jpeg artifacts normally imgmin has always given me images that I can't tell the difference between at 100% zoom only if I zoom in waymore than that can I begin to notice it.

imgmin's "optimzied" file that it says is nearly identical to the original image when it clearly is not as there is a crapton of compression artifacts throughout the image.

original file quailty was 90 non-progresive non optimized even huffman optimied via jpegtran

Debian's pngquant is 13 years old and has very poor image quality compared to modern pngquant.

Consider using packages from http://pngquant.org

The new pngquant also has --quality argument which makes it automatically select number of colors required to achieve given quality (takes unit similar to JPEG's quality, optimizes for mean square error).

Something I'd like to have it do is to just show space saved, or be able to show nothing. As it is now, it's filling up the terminal with information that I'd rather not see. So if there is a way to make it so that it just shows "space saved", or "skipped". That'd be amazing.

Many old black-and-white photos are erroneously of the TrueColor type -- this may also open the door for specific b&w optimizations. These should be detected and converted to Grayscale where possible. Example:

http://upload.wikimedia.org/wikipedia/commons/thumb/1/1b/Abraham_Lincoln_November_1863.jpg/220px-Abraham_Lincoln_November_1863.jpg

At the bottom

imgmin

$ git clone [email protected]:rflynn/imgmin.git
$ cd imgmin
$ make
$ sudo make install

Shouldn't it be autoconf?

Running the following script that I've written to do the "recursive" thing does not work how he said it should. Nor does /dev/null work.

#/bin/bash
cd $1;
find -iname "*.jpg" >> tmp;
sed -i "s/.\///" tmp;

FILENAME=tmp;
while read line
do
      imgmin "$line" "$line" | grep -o 'saved:.*';
      jhead -purejpg "$line";
done < $FILENAME

Please note I've also tried using it with > /dev/null and it doesn't work either, so thus it's doing something weird.

I'm on debian with the following uname.

Linux sager-laptop 2.6.39-bpo.2-amd64 #1 SMP Tue Jul 26 10:35:23 UTC 2011 x86_64 GNU/Linux

I compiled it for 64bits using the following gcc commands.

-march=core2 -O3

other than that it was just a straight compile. I have no idea why it's acting so weird and not working with the utilities as it should.

I know that this is probably not something that you were thinking of when you first made it, but not being able to go through and run it on a directory of images that i have is a major annoyance. I know it's made for websites, but I wouldn't mind making my photos that I have on my computer take up less space.

Anyway this is an awesome utility from it's description, and also running it.

Re #34 I'd prefer not to build Apache module at all, but I don't see the option for that in ./configure --help and the generated makefile doesn't have imgmin target (which I hoped would just build the commandline version):

$ make imgmin
make: *** No rule to make target `imgmin'.  Stop.

This is like JPEGmini w/o infrastructure.
Thank you!

Could you explain what are pngnq pngcrush pngquant for in case of JPEGs?

Could you create tagged releases?
https://help.github.com/articles/creating-releases/