Struct Zensim 

pub struct Zensim { /* private fields */ }

Metric configuration. Methods on this struct are the primary API.

use zensim::{Zensim, ZensimProfile, RgbSlice};
let z = Zensim::new(ZensimProfile::latest());
let source = RgbSlice::new(&src, 8, 8);
let distorted = RgbSlice::new(&dst, 8, 8);
let result = z.compute(&source, &distorted).unwrap();
println!("{}: {:.2}", result.profile(), result.score());

Implementations

impl Zensim

pub fn compute_with_ref_and_diffmap( &self, precomputed: &PrecomputedReference, distorted: &impl ImageSource, options: impl Into<DiffmapOptions>, ) -> Result<DiffmapResult, ZensimError>

Compare with precomputed reference and return both score and per-pixel error map.

The diffmap fuses SSIM error maps from all pyramid scales, weighted by the profile’s trained weights. Coarser scales are upsampled to full resolution and blended. The zensim score uses the identical multi-scale pipeline.

Use case

Encoder quantization loops: precompute the reference once, then in each iteration call this to get both a global quality score (for convergence) and a spatial error map (for per-block quant field adjustment).

Errors

Returns ZensimError::ImageTooSmall if dimensions < 8×8.

pub fn compute_with_ref_and_diffmap_linear_planar( &self, precomputed: &PrecomputedReference, planes: [&[f32]; 3], width: usize, height: usize, stride: usize, options: impl Into<DiffmapOptions>, ) -> Result<DiffmapResult, ZensimError>

Compare planar linear RGB f32 against a precomputed reference, producing both a score and a per-pixel error map.

planes are [R, G, B], each with at least stride * height elements. stride is the number of f32 elements per row (≥ width).

This avoids the interleave-to-RGBA overhead when the caller already has separate channel buffers in linear light (e.g., from an encoder’s reconstruction pipeline).

Errors

Returns ZensimError::ImageTooSmall if dimensions < 8×8.

pub fn compute_with_diffmap( &self, source: &impl ImageSource, distorted: &impl ImageSource, options: impl Into<DiffmapOptions>, ) -> Result<DiffmapResult, ZensimError>

Compare two images and return both score and per-pixel error map.

Convenience method that handles precomputation internally. For iterative use (encoder loops), prefer precompute_reference + compute_with_ref_and_diffmap to avoid re-converting the reference each time.

Errors

Returns ZensimError if dimensions are mismatched or too small.

impl Zensim

pub fn new(profile: ZensimProfile) -> Self

Create a new Zensim with the given profile. Parallel by default.

pub fn with_parallel(self, parallel: bool) -> Self

Enable or disable multi-threaded computation (rayon). Default: true.

pub fn profile(&self) -> ZensimProfile

Current profile.

pub fn parallel(&self) -> bool

Whether multi-threaded computation is enabled.

pub fn compute( &self, source: &impl ImageSource, distorted: &impl ImageSource, ) -> Result<ZensimResult, ZensimError>

Compare source and distorted images.

Errors

Returns ZensimError if dimensions are mismatched or too small.

pub fn precompute_reference( &self, source: &impl ImageSource, ) -> Result<PrecomputedReference, ZensimError>

Pre-compute reference image data for batch comparison.

Errors

Returns ZensimError::ImageTooSmall if dimensions < 8×8.

pub fn compute_with_ref( &self, precomputed: &PrecomputedReference, distorted: &impl ImageSource, ) -> Result<ZensimResult, ZensimError>

Compare a distorted image against a precomputed reference.

Errors

Returns ZensimError::ImageTooSmall if dimensions < 8×8.

pub fn precompute_reference_linear_planar( &self, planes: [&[f32]; 3], width: usize, height: usize, stride: usize, ) -> Result<PrecomputedReference, ZensimError>

Precompute reference from planar linear RGB f32 data.

planes are [R, G, B], each with at least stride * height elements. stride is the number of f32 elements per row (≥ width; may be larger for padded buffers like the encoder’s padded_width).

This avoids the interleave-to-RGBA overhead when the caller already has separate channel buffers in linear light.

Errors

Returns ZensimError::ImageTooSmall if dimensions < 8×8.

Trait Implementations

impl Clone for Zensim

fn clone(&self) -> Zensim

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Zensim

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.