CharLS is a C++ implementation of the JPEG-LS standard for lossless and near-lossless image compression and decompression. JPEG-LS is a low-complexity image compression standard that matches JPEG 2000 compression ratios.
- C++17 library implementation with a binary C interface for maximum interoperability.
- Supports Windows, Linux and macOS on x86, x64, arm, arm64, mips, s390x and ppc64le.
- Adapters for .NET, JavaScript (WebAssembly) and Python applications available.
- Excellent compression and decompression performance.
JPEG-LS (ISO/IEC 14495-1:1999 / ITU-T.87) is an image compression standard derived from the Hewlett Packard LOCO algorithm. JPEG-LS has low complexity (meaning fast compression) and high compression ratios, similar to the JPEG 2000 lossless ratios. JPEG-LS is more similar to the old Lossless JPEG than to JPEG 2000, but interestingly the two different techniques result in vastly different performance characteristics.
Wikipedia on lossless JPEG and JPEG-LS: https://en.wikipedia.org/wiki/Lossless_JPEG
Tip: ITU makes their version of the JPEG-LS standard (ITU-T.87) freely available for download. The text is identical with the ISO version.
This project's goal is to provide a full implementation of the ISO/IEC 14495-1:1999, "Lossless and near-lossless compression of continuous-tone still images: Baseline" standard. This library is written from scratch in portable C++. The main branch uses C++17. The 2.x branch is maintained in C++14. All mainstream JPEG-LS features are implemented by this library. According to preliminary test results published on https://imagecompression.info/gralic, CharLS is about twice as fast as the original HP code, and beats both JPEG-XR and JPEG 2000 by a factor 3.
The following JPEG-LS options are not supported by the CharLS implementation. Most of these options are rarely used in practice.
- No support to encode JPEG restart markers
Decoding is supported, but no recovery mechanisme is implemented for corrupted JPEG-LS files. - No support for sub-sampled scans. Sub-sampling is a lossly encoding mechanism and not used in lossless scenarios.
- No support for multi component frames with mixed component counts in a single scan. While technical possible all known JPEG-LS codecs put multi-component (color) images in a single scan or in multiple scans, but not use a mix of these in one file.
- No support to encode\decode images with the height defined after the first scan (DNL marker).
- No support for JPEG-LS mapping tables (palette).
- No support for point transform. Point transform is a lossly encoding mechanism and not used in lossless scenarios.
After releasing the original baseline JPEG-LS standard ISO 14495-1:1999, ISO released an extension to the JPEG-LS standard called ISO/IEC 14495-2:2003: "Lossless and near-lossless compression of continuous-tone still images: Extensions". Currently CharLS doesn't support these extensions.
The code is regularly compiled/tested on Windows and 64 bit Linux. Additionally, the code has been successfully tested on Linux Intel/AMD 32/64 bit (slackware, debian, gentoo), Solaris SPARC systems, Intel based Macs and Windows CE (ARM CPU, emulated), where the less common compilers may require minor code edits. It leverages C++ language features (templates, traits) to create optimized code, which generally perform best with recent compilers. If you compile with GCC, 64 bit code performs substantially better.
With vcpkg on Windows
PS> vcpkg install charls charls:x64-windowsWith vcpkg on Linux or macOS
~/$ ./vcpkg install charlsFor other platforms, more install options, how to build from source, and more, take a look at the documentation.
Once you have the library, the sample folder provides some code samples to get you started.
This repository is provided as source code, and specifically does not offer binary releases. Instead, it is encouraged to either “live at head” (build from the latest version of or, if necessary, build against a known, supported branch, known as a Long Term Support (LTS) branch. Support for older compiler versions will be phased out, 5 years from the moment that a newer version of that compiler has become available. The same applies to the minimal required C++ language version.
Before any major breaking change in the API and/or ABI a branch will be created from main to freeze that snapshot as LTS branch.
- CharLS.Native .NET - a .NET adapter assembly for CharLS: NuGet package
- JPEG-LS WIC codec - Windows Imaging Component (WIC) codec for JPEG-LS .jls files
- charls-js - WebAssembly build of CharLS: Demo
- JPEG-LS plugin for Python Pillow - JPEG-LS plugin for Python Pillow: Pip package
CharLS is being used by GDCM DICOM toolkit, thanks for Mathieu Malaterre for getting CharLS started on Linux. Kato Kanryu wrote an initial version of the color transforms and the DIB output format code, for an irfanview plugin using CharLS.
The code in this project is available through a BSD style license, allowing use of the code in commercial closed source applications if you wish. All the code in this project is written from scratch, and not based on other JPEG-LS implementations. Be aware that Hewlett Packard claims to own patents that apply to JPEG-LS implementations, but they license it for free for conformant JPEG-LS implementations. Some of these patents may already have expired in your country. Read more at https://hpl.hp.com/loco/ before you use this if you use this code for commercial purposes.