Kornia: kornia-io

Image and Video Input/Output library for the Kornia ecosystem.

🚀 Overview

kornia-io provides high-performance utilities for reading and writing images and video streams. It abstracts over common formats and libraries to provide a unified, type-safe API for getting visual data into your Rust applications. It supports standard image formats (JPEG, PNG, TIFF) and integrates with GStreamer and V4L2 for advanced video capture. The generic read_image/decode_image/write_image helpers exposed by the Python bindings (kornia-py) are Python-only; Rust users should call the typed per-format functions (jpeg, png, tiff) in this crate directly.

🔑 Key Features

• Image I/O: Read and write support for JPEG, PNG, and TIFF formats.
• In-memory encode + decode: Symmetric encode_image_{jpeg,png}_* ↔ decode_image_{jpeg,png}_* for round-tripping pixels through Vec<u8> without touching disk — including lossless PNG-16 (encode_image_png_gray16) for depth maps.
• TurboJPEG Support: Optional integration with turbojpeg for high-performance JPEG encoding and decoding.
• Video Capture (GStreamer): Access generic video streams (files, IP cameras, webcams) via GStreamer integration.
• Camera Access (V4L2): Direct low-latency access to V4L2 devices on Linux.
• Type Integration: Returns kornia-image structs directly, ensuring seamless interoperability with the rest of the ecosystem.

📦 Installation

Add the following to your Cargo.toml. Select features based on your needs:

[dependencies]
kornia-io = { version = "0.1.0", features = ["turbojpeg", "gstreamer"] }

🛠️ Usage

Reading an Image

use kornia_io::functional as F;
use kornia_image::color_spaces::Rgb8;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Read an image from disk
    // automatically detects format (JPEG, PNG, TIFF)
    let image = F::read_image_any_rgb8("path/to/image.jpg")?;

    println!("Image loaded: {}x{}", image.width(), image.height());

    Ok(())
}

Video Capture (GStreamer)

Requires gstreamer and v4l features.

use kornia_io::gstreamer::{CameraCapture, V4L2CameraConfig};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create a camera capture object
    let mut capture = CameraCapture::new(
      &V4L2CameraConfig::new().with_camera_id(0).with_fps(30)
    )?;

    // Start the capture pipeline
    capture.start()?;

    // Grab a frame
    if let Some(img) = capture.grab_rgb8()? {
        println!("Image captured: {:?}", img.size());
    }

    Ok(())
}

🧩 Modules

• error: Error types for the I/O module.
• functional: High-level helper functions like read_image_any_rgb8.
• jpeg / png / tiff: Format-specific implementations.
• jpegturbo: (Feature: turbojpeg) High-performance JPEG encoding and decoding via TurboJPEG.
• gstreamer: (Feature: gstreamer) Video capture and streaming via GStreamer.
• v4l: (Feature: v4l, Linux only) Direct Video4Linux2 camera access.
• fps_counter: Utilities for measuring frame rates.

💡 Related Examples

You can find comprehensive examples in the examples folder of the repository:

• rtspcam: RTSP camera streaming using GStreamer.
• v4l: Video capture using V4L2.
• video_player: Simple video player example.
• video_write: Video recording and writing.
• foxglove: Integration with Foxglove Studio for visualization.
• exif_auto_orient: Compare raw JPEG decode vs EXIF auto-oriented output.

🤝 Contributing

Contributions are welcome! This crate is part of the Kornia workspace. Please refer to the main repository for contribution guidelines.

📄 License

This crate is licensed under the Apache-2.0 License.