A collection of scaling algorithms for "pixel-art", combined into a single, stand-alone tool, using a common, general-purpose programming language.
pixelscaler algo input.bmp [output.bmp]
The input filename is given as the second argument. The input file must be in Bitmap (BMP/BMP3) format, Version 3, 24 bits per pixel. Generated output uses the same bitmap format.
The third argument is optional; if it is omitted, the
output file will be named output.bmp.
The first argument selects the scaling algorithm to use, it must
be one of: block2, block3, scale2x, scale2xSFX, scale3x,
scale3xSFX, hq2xA, hq2xB, hq3xA, hq3xB, superXBR.
Other file formats must be converted to BMP3 first; many tools (like
ImageMagick or the Gimp) can do that. Just be sure to specify 24bit
colordepth. For example, using ImageMagick, you might use:
convert input.gif -type truecolor input.bmp3.
git clone https://github.com/janert/pixelscalers.git
cd pixelscalers/src
make
This tool combines implementations of several of the well-known scaling algorithms for pixel art.
block2: Each input pixel is expanded into a 2x2 block; no interpolation.block3: Each input pixel is expanded into a 3x3 block; no interpolation.scale2x: The Scale2x algorithm, 2x magnification.scale2xSFX: The improvedscale2xalgorithm by Sp00kyFox, 2x magnification.scale3x:The Scale2x algorithm, 3x magnification.scale3xSFX: The improvedscale3xalgorithm by Sp00kyFox, 3x magnification.hq2xA: The Hqx algorithm, optimized for simple graphs, 2x magnification.hq2xB: The Hqx algorithm, optimized for complex graphs, 2x magnification.hq3xA: The Hqx algorithm, optimized for simple graphs, 3x magnification.hq3xB: The Hqx algorithm, optimized for complex graphs, 3x magnification.superXBR: The Super xBR algorithm, 2x magnification.
Not included is the 2×SaI algorithm. Maybe I will add it at some point.
For some algorithms (in particular the
SFX
versions and the
XBR
algorithm),
the primary documentation seem to be forum discussions. Screenshots of those
are included as PDFs in the docs directory, to preserve them for reference.
These images are BMP files; if they don't show up, try a different browser.
Original  |
||
block2  |
block3  |
|
scale2x  |
scale3x  |
|
scale2xSFX  |
scale3xSFX  |
|
hq2xA  |
hq3xA  |
|
hq2xB  |
hq3xB  |
|
superXBR  |
-
IO using standard input/output is not supported, so the tool can unfortunately not be used as a filter.
-
The size of integer types used in the implementation is not always consistent. The implementation may fail for very large images (more than 16k pixels along one edge) because of insufficient integer range.
-
Since they were not intended for real-time processing, no effort has been made to optimize the execution time of the algorithm implementations.
-
The code layout has not been unified across the various original implementations.
This is largely derivative work. For a personal project, I wanted to use and compare the effects of a variety of published algorithms, but existing implementations seemed to be somewhat scattered about. Some algorithms appear to have only been discussed on forums, with no implementations in a general-purpose language easily accessible, or only available via the Wayback Machine. So I thought it might be useful to collect them in one place and provide a consistent wrapper around them.
This tool incorporates the works of others. If you are one of the original authors and you think I am infringing on your rights, let me know, so we can work it out.
-
The implementation of the ScaleNx algorithms follows the descriptions here, and here.
-
The implementation of the Hqx algorithms, as well as the bitmap handling are largely based on this.
-
The implementation of the Super xBR algorithm is taken from here. Also see the discussion.
-
It's been a long time since I wrote a Makefile; I found this page a helpful reminder.
Copyright (c) 2022, Philipp K. Janert. All rights reserved.
Unless otherwise stated in individual files, the contents of this repository is licensed under the MIT-style license found in the LICENSE file in this directory.