Lucas Canade Optical Flow and Shi-Tomasi feature detection on Rust

High-performance Rust implementation of Lucas-Kanade optical flow and Shi-Tomasi feature detection, optimized for real-time applications and WebAssembly (Wasm) compatibility.

Features

• 🔍 Efficient feature point detection using Shi-Tomasi
• 🖼️ Integration with image and imageproc crates
• 🌐 WebAssembly (Wasm) compatible

Usage

Add to your Cargo.toml:

[dependencies]

optical-flow-lk = "0.1"

Basic example:

use image::{open, GrayImage, Rgba};
use optical_flow_lk::{build_pyramid, calc_optical_flow, good_features_to_track};

let prev_frame: GrayImage = open("examples/input1.png").unwrap().clone().into_luma8();
let next_frame: GrayImage = open("examples/input2.png").unwrap().clone().into_luma8();

let prev_frame_pyr = build_pyramid(&prev_frame, 4);
let next_frame_pyr = build_pyramid(&next_frame, 4);

let mut points = good_features_to_track(&prev_frame, 0.1, 5);
points.truncate(100);
let prev_points: Vec<(f32, f32)> = points.iter().map(|&x| (x.0 as f32, x.1 as f32)).collect();

let next_points = calc_optical_flow(&prev_frame_pyr, &next_frame_pyr, &prev_points, 21, 30);

Live demo

A browser demo runs the tracker entirely client-side in WebAssembly: point your phone's camera at a scene and tap to drop points (or hit Auto to detect Shi-Tomasi corners) and watch them ride the optical flow.

▶ lk-demo.jt3.ru

Source and build instructions are in web-demo/.

WebAssembly

The Scharr gradient kernel has a hand-written simd128 path that is selected automatically on wasm32 — but only when the target is built with SIMD enabled, since WASM has no runtime feature detection. The bundled .cargo/config.toml sets this for you:

[target.wasm32-unknown-unknown]

rustflags = ["-C", "target-feature=+simd128"]

If you build from a different working directory (so that config is not picked up), pass it yourself:

RUSTFLAGS="-C target-feature=+simd128" cargo build --release --target wasm32-unknown-unknown

Without +simd128 the crate still works, falling back to a scalar gradient loop. For production WASM, also run wasm-opt -O3 on the output. On a 640×480 per-frame track step, the simd128 build plus the bounds-check-free bilinear sampler is roughly 2× faster than the scalar build in V8 (Node).