Jpeg Reader mojo-native library

Community Showcase
1

Hello everyone!

I wanted to share my open-source project called mojo-image-reader. It is a lightweight and dependency-free JPEG decoder written entirely in the Mojo v1.1.

Purpose: Built as a portfolio showcase and a useful tool for the Mojo ecosystem.

Features:

  • JPEG Support: Baseline (8-bit and 12-bit precision SOF1), Progressive JPEG decoding, and full Chroma Subsampling support (4:4:4, 4:2:2, 4:2:0, and 4:2:0v).

  • Performance & Memory Optimizations: Built with raw pointer arithmetic to bypass bounds-checking.

  • Flexible Export & Interoperability: PPM Exporter (8-bit and 12/16-bit integer color depths), PFM Exporter (32-bit Float HDR), and a NumPy Bridge for zero-copy data exchange.

Project link: https://github.com/AlexALX/mojo-image-reader

Want to know your opinion and does it have any sense to continue and write things listed in the roadmap (BMP, PNG, GIF support)? I would appreciate any feedback or suggestions!

Thanks!

2

Pushed new updated, so now it supports 1-component jpeg (grayscale), proper restart markers support and optimized with SIMD instructions, made it work in x5 times faster.

3

Did major updates and finished the roadmap!

The library now supports parsing JPG, PNG, GIF, and BMP, with export capabilities to PPM, PGM, PFM, and PAM formats.

Key highlights:

  • GIF: Frame sequence extraction with proper disposal handling.

  • Precision: Export to 8, 12, or 16-bit files.

  • Performance: Hardware-adaptive SIMD acceleration where applicable + zero-alloc memory optimizations.

4

Also created a benchmark and sample repository: mojo-image-reader-bench, complete with performance charts comparing it against Python’s Pillow.

benchmark_comparison
benchmark_comparison2400×1800 406 KB

Of course, there’s still room to optimize, and I might improve it more later, but the results so far are already very good!