Module simd 

SIMD-accelerated image processing operations

Provides optimized implementations for common image operations using AVX2, AVX-512, and ARM NEON intrinsics.

Functions

fast_area_resize

Ultra-fast area average downscaling for preprocessing Best for large images being scaled down significantly

parallel_simd_resize

Parallel SIMD resize for large images - splits work across threads

simd_grayscale

Convert RGBA image to grayscale using optimized SIMD operations

simd_normalize

Normalize f32 tensor data using SIMD

simd_resize_bilinear

Fast bilinear resize using SIMD - optimized for preprocessing This is significantly faster than the image crate’s resize for typical OCR sizes

simd_threshold

Apply threshold to grayscale image using SIMD