Skip to content
/ piliq Public

Lightweight Python PIL-libimagequant/pngquant interface with autonomous lib look-up.

License

Notifications You must be signed in to change notification settings

cubicibo/piliq

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

piliq

PIL-LibImageQuant basic and lightweight interface to quantize images. An elementary auto-DLL look-up and loading mechanism is included.

How it works

A PILIQ instance is loaded by providing a string, which may be a path to an executable, a dll, or a command name. This operation arms the library and enables the caller to quantize RGBA images. The exposed interface is kept to a bare minimum and limited to selected parameters mutually exposed by both pngquant and libimagequant.

Examples

For pngquant:

piliq = PILIQ("path_to_pngquant.exe")

For libimagequant:

piliq = PILIQ("path_to_libimagequant.dll")

In either case, this causes the internal wrappers to be initialised once and for all. All subsequent calls to PILIQ constructor will return the library initialised before.

Additionally, PILIQ embeds an elementary autonomous look-up mechanism to find pngquant or libimagequant.

  • If you are on Windows x86_64, PILIQ will always manage to at least bind to the provided libimagequant dll.
  • Else, PILIQ will attempt to acccess pngquant as a subprocess, assuming it may be in your environment path (quite common for users on Linux or macOS).
  • If none of the above stands true, PILIQ looks up libimagequant in known folders collecting dynamic libraries.

To quantize:

img_rgba = Image.open(...).convert('RGB')
#Quantize on a maximum of 223 colors/clusters.
img_p = piliq.quantize(img_rgba, 223)
img_p.save(...)

If you wish to work with numpy or bitmaps and palettes, you may set piliq.return_pil = False. This will change the return type of quantize() to a tuple containing:

palette, bitmap = img_p

Options

Set the default maximum quality acceptable. Quality is a libimagequant metric bounded between 1 and 100, incl. At 100, libimagequant may always use the highest number of colors, while visually identical results using much less could be obtained in qualities rangin from the 80s to 90s.

piliq.set_quality(quality: int)

Set the default speed. Speed is a notion defined by libimagequant and it refers to the algorithms used in the quantization process. The modes ranges from 1 to 10, incl.

piliq.set_speed(speed: int)

Set the default maximum number of colors that the result may feature. The argument of the quantize function overrides the default if provided. Paletted images have a maximum of 256 entries.

piliq.set_default_max_colors(colors: int)

Set the dithering level of the output. Libimagequant dithers the quantization results, picking close by colors in the output palette. This does not change the computation speed nor may have any visible impact depending of the settings. The value is a float within [0.0, 1.0], incl.

piliq.set_dithering_level(dithering_level: float)

Limits

  • libimagequant is not supported on macOS, only pngquant.
  • The internal look-up is rather limited and simple, it may fail everytime. Window python package embeds the x86_64 dll of libimagequant which will auto-load if not provided.
  • The provided x86_64 dll is not embedded in frozen "onefile" executables produced via PyInstaller or similar app packagers.

About

Lightweight Python PIL-libimagequant/pngquant interface with autonomous lib look-up.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages