#webp #png #jpeg #gif #input-file #filesize

libcaesium

A lossy/lossless image compression library

1 unstable release

0.17.0 Oct 11, 2024

#260 in Images

Apache-2.0 and maybe GPL-3.0-or-later

73KB
1.5K SLoC

libcaesium Rust

Libcaesium is a simple library performing JPEG, PNG, WebP and GIF (experimental) compression/optimization written in Rust, with a C interface.

[!WARNING] starting from v0.6.0 the library is written in Rust and no longer in C. There's a C interface, but it's not backward compatible with the <0.6.0.

Usage example

Libcaesium exposes two functions, auto-detecting the input file type

use caesium::parameters::CSParameters;
use caesium::compress;

let mut parameters = CSParameters::new();
parameters.keep_metadata = true;
parameters.jpeg.quality = 60;

let success = compress(input, output, &parameters).is_ok();

Compilation

Compilation is available for all supported platforms: Windows, macOS and Linux.

[!NOTE] if you don't use the --release flag, the PNG optimizations can take a very long time to complete, especially using the zopfli algorithm.

cargo build --release

The result will be a dynamic library usable by external applications through its C interface.

Usage in C

Libcaesium exposes two C functions, auto-detecting the input file type:

Based on quality values

pub unsafe extern "C" fn c_compress(
    input_path: *const c_char,
    output_path: *const c_char,
    params: CCSParameters
) -> CCSResult

Parameters

  • input_path - input file path (full filename)
  • output_path - output file path (full filename)
  • parameters - options struct, containing compression parameters (see below)

Return

A CCSResult struct

#[repr(C)]
pub struct CCSResult {
    pub success: bool,
    pub error_message: *const c_char,
}

If success is true the compression process ended successfully and error_message will be empty.
On failure, the error_message will be filled with a string containing a brief explanation of the error.

Based on output size

pub unsafe extern "C" fn c_compress_to_size(
    input_path: *const c_char,
    output_path: *const c_char,
    params: CCSParameters,
    max_output_size: usize,
) -> CCSResult

Parameters

  • input_path - input file path (full filename)
  • output_path - output file path (full filename)
  • parameters - options struct, containing compression parameters (see below)
  • max_output_size - the maximum output size, in bytes

Return

A CCSResult struct

#[repr(C)]
pub struct CCSResult {
    pub success: bool,
    pub error_message: *const c_char,
}

If success is true the compression process ended successfully and error_message will be empty.
On failure, the error_message will be filled with a string containing a brief explanation of the error.

Compression options

The C options struct is slightly different from the Rust one:

#[repr(C)]
pub struct CCSParameters {
    pub keep_metadata: bool,
    pub jpeg_quality: u32,
    pub jpeg_chroma_subsampling: u32,
    pub png_quality: u32,
    pub png_optimization_level: u32,
    pub png_force_zopfli: bool,
    pub gif_quality: u32,
    pub webp_quality: u32,
    pub tiff_compression: u32,
    pub tiff_deflate_level: u32,
    pub optimize: bool,
    pub width: u32,
    pub height: u32,
}

The option description is the same as the Rust counterpart.
Valid values for jpeg_chroma_subsampling are [444, 422, 420, 411]. Any other value will be ignored and will be used the default option.
Valid values for tiff_compression are [0 (Uncompressed), 1 (Lzw), 2 (Deflate), 3 (Packbits)]. Any other value will be ignored and 0 will be used.
Valid values for tiff_deflate_level are [1 (Fast), 6 (Balanced), 9 (Best)]. Any other value will be ignored and Best will be used.

Compression vs Optimization

JPEG is a lossy format: that means you will always lose some information after each compression. So, compressing a file with 100 quality for 10 times will result in an always different image, even though you can't really see the difference. Libcaesium also supports optimization, by setting the quality to 0. This performs a lossless process, resulting in the same image, but with a smaller size (10-12% usually).
GIF optimization is possible, but currently not supported. WebP's optimization is also possible, but it will probably result in a bigger output file as it's well suited to losslessly convert from PNG or JPEG.

Dependencies

~9–18MB
~198K SLoC