sci-gaussfilt-rs

A Rust implementation of the 1D Gaussian filter that exactly matches the behavior of scipy.ndimage.gaussian_filter1d.

This crate provides a high-performance, pure Rust implementation of the Gaussian filter for 1D arrays, supporting various boundary conditions and floating-point types (f32, f64). It is designed to be a drop-in replacement for SciPy's implementation in Rust applications.

Features

• SciPy Parity: Matches scipy.ndimage.gaussian_filter1d output within floating-point tolerance (1e-6 or better).
• Boundary Modes: Supports Reflect, Nearest, and Constant padding modes.
• Generic: Works with f32 and f64 data types.
• Performance: Optimized convolution with pre-calculated kernels and padding.

Installation

Add this to your Cargo.toml:

[dependencies]
sci-gaussfilt-rs = "0.1.0"

Usage

use sci_gaussfilt_rs::{gaussian_filter1d, types::BoundaryMode};

fn main() {
    let data = vec![1.0, 2.0, 3.0, 4.0, 5.0];
    let sigma = 1.0;
    let truncate = 4.0;
    
    // Apply Gaussian filter with Reflect boundary mode
    let result = gaussian_filter1d(&data, sigma, truncate, BoundaryMode::Reflect).unwrap();
    
    println!("Filtered data: {:?}", result);
}

Supported Boundary Modes

• BoundaryMode::Reflect: (Default in SciPy) d c b a | a b c d | d c b a
• BoundaryMode::Nearest: a a a a | a b c d | d d d d
• BoundaryMode::Constant(value): v v v v | a b c d | v v v v

License

MIT