pngpetite

A script to reduce the size of PNG images significantly. The code is based on Crush by Christopher Simpkins and has been adapted to Linux by refactoring it. As the code base shows significant differences the name has also been changed.

Prerequisites

bc
file
pngquant
zopflipng
butteraugli (optional)

Usage

Usage:
  pngpetite <options> <png-image> ...

Options:
  -h | --help         This message
  -r | --replace      Replace file with processed one
                      Implies -d
  -d | --dest <path>  Place images in given path
                      Implies -r
  -S | --no-stats     Don't print stats
  -q | --quiet        Don't print stats and progress
  -V | --version      Print version and exit

Description:
  Insane(ly slow but wicked good) PNG image optimization
  Built with pngquant and zopflipng optimizers

Examples

Default usage. Yields happy-sun-pngpetite.png

pngpetit happy-sun.png
Original (KB)   Stage1 (KB)    Final (KB)    Saved (KB)       Saved %    Time (sec)  Image Name
------------- ------------- ------------- ------------- ------------- -------------  -------------
       25.639        16.186        13.857        11.781        45.951        22.204  happy-sun.png

Replace existing image. Yields happy-sun.png

pngpetit --replace happy-sun.png
Original (KB)   Stage1 (KB)    Final (KB)    Saved (KB)       Saved %    Time (sec)  Image Name
------------- ------------- ------------- ------------- ------------- -------------  -------------
       25.639        16.186        13.857        11.781        45.951        22.204  happy-sun.png

Place processed image in new directory. Yields pngpetite/happy-sun.png

pngpetit --dest pngpetite happy-sun.png
Original (KB)   Stage1 (KB)    Final (KB)    Saved (KB)       Saved %    Time (sec)  Image Name
------------- ------------- ------------- ------------- ------------- -------------  -------------
       25.639        16.186        13.857        11.781        45.951        22.204  happy-sun.png

Process all PNG images in directory originals and place them in directory pngpetite.

pngpetite --dest pngpetite original/*png
Original (KB)   Stage1 (KB)    Final (KB)    Saved (KB)       Saved %    Time (sec)  Image Name
------------- ------------- ------------- ------------- ------------- -------------  -------------
      695.867       695.867       658.104        37.764         5.427        20.190  original/beach.png
       25.639        16.186        13.857        11.781        45.951        22.197  original/happy-sun.png
       54.101        29.885        27.244        26.856        49.642         8.097  original/shinkansen-n700.png

Docker usage

docker run \
  --rm \
  --tty \
  --volume "$(pwd)":"$(pwd)" \
  --workdir "$(pwd)" \
  --user "$(id -u):$(id -g)" \
  uroesch/pngpetite <PNG image> [<PNG image> ...]

Or simply use the docker-pngpetite wrapper script found in the bin directory.

bin/docker-pngpetite <PNG image> [<PNG image> ...]

Visual Examples

pngpetite is not suited for every type of picture. Especially photographic images may not yield high savings. To visualize the differences between the original image and the optimized one, google’s butteraugli tools is used.

ℹ️	The values presented below were done on an Ubuntu 20.10 machine with: `pngquant 2.12.0 (January 2018)` `zopfli 1.0.3` Your milage may vary.

Photographs

Beach Sunset

Table 1. Processing Results

Original (KB)	Stage1 (KB)	Final (KB)	Saved (KB)	Saved %	Butteraugli Score
443.642	164.648	150.155	293.486	66.154	2.676426

Noticeable is that at stage 1 pngquant shrinks the image by 62% in older versions photographic pictures did only yield a reduction in size. At stage 2 zopflipng is able to squeeze another 4% from its compression algorithm. Comparing the two images with butteraugli shows a large differences especially in the sunset section of the image.

Original

Optimized

Heatmap

Grey Cat

Table 2. Processing Results

Original (KB)	Stage1 (KB)	Final (KB)	Saved (KB)	Saved %	Butteraugli Score
567.165	204.628	192.016	375.149	66.145	3.551921

The heatmap shows clearly where the biggest differences occur during the processing. Notable are they eyes and the background. The fur is very similar to the originial image.