Struct ImageFeatureSelector 

pub struct ImageFeatureSelector<State = Untrained> { /* private fields */ }

Image feature selection using spatial correlation and frequency analysis

This selector analyzes image features represented as flattened pixel matrices or extracted feature vectors and applies image-specific selection criteria:

• Spatial correlation analysis for neighboring pixels
• Frequency domain analysis using variance as a proxy
• Texture analysis using local variance measurements
• Combined scoring with target correlation

Input Format

The input matrix X should be structured as:

• Rows: Images
• Columns: Features (pixels, extracted features, etc.)
• Values: Pixel intensities, feature values, or derived measurements

Examples

use sklears_feature_selection::domain_specific::image_features::ImageFeatureSelector;
use sklears_core::traits::{Fit, Transform};
use scirs2_core::ndarray::{Array1, Array2};

let selector = ImageFeatureSelector::new()
    .include_spatial(true)     // Enable spatial correlation analysis
    .include_frequency(true)   // Enable frequency domain analysis
    .include_texture(true)     // Enable texture analysis
    .spatial_threshold(0.15)   // Threshold for spatial features
    .k(Some(100));             // Select top 100 features

let x = Array2::zeros((50, 784)); // 50 images, 784 pixels (28x28)
let y = Array1::zeros(50);         // Image labels

let fitted_selector = selector.fit(&x, &y)?;
let transformed_x = fitted_selector.transform(&x)?;

Implementations

impl ImageFeatureSelector<Untrained>

pub fn new() -> Self

Create a new image feature selector with default parameters

Default configuration:

• include_spatial: true
• include_frequency: true
• include_texture: true
• spatial_threshold: 0.1
• k: None (use threshold-based selection)

pub fn include_spatial(self, include_spatial: bool) -> Self

Enable or disable spatial correlation analysis

When enabled, the selector computes correlations between features and the target variable, emphasizing spatial relationships.

pub fn include_frequency(self, include_frequency: bool) -> Self

Enable or disable frequency domain analysis

When enabled, the selector analyzes frequency content using variance as a proxy for spectral energy, combined with target correlation.

pub fn include_texture(self, include_texture: bool) -> Self

Enable or disable texture analysis

When enabled, the selector computes local variance measurements to identify texture-rich regions that correlate with the target.

pub fn spatial_threshold(self, threshold: f64) -> Self

Set the threshold for spatial correlation selection

Features with combined scores below this threshold will be filtered out (when not using k-based selection).

pub fn k(self, k: Option<usize>) -> Self

Set the number of top features to select

When set to Some(k), selects the top k features by combined score. When set to None, uses threshold-based selection.

impl ImageFeatureSelector<Trained>

pub fn spatial_scores(&self) -> Option<&Array1<Float>>

Get the spatial correlation scores (if spatial analysis was enabled)

Returns None if spatial analysis was not enabled during fitting.

pub fn frequency_scores(&self) -> Option<&Array1<Float>>

Get the frequency domain scores (if frequency analysis was enabled)

Returns None if frequency analysis was not enabled during fitting.

pub fn texture_scores(&self) -> Option<&Array1<Float>>

Get the texture scores (if texture analysis was enabled)

Returns None if texture analysis was not enabled during fitting.

pub fn selected_features(&self) -> &[usize]

Get the indices of selected features

pub fn n_features_selected(&self) -> usize

Get the number of selected features

pub fn feature_summary( &self, ) -> Vec<(usize, Option<Float>, Option<Float>, Option<Float>)>

Get a summary of feature scores across all analysis types

Returns a vector of tuples containing (feature_index, spatial_score, frequency_score, texture_score) for all selected features. Scores are None if the corresponding analysis was not enabled.

Trait Implementations

impl<State: Clone> Clone for ImageFeatureSelector<State>

fn clone(&self) -> ImageFeatureSelector<State>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<State: Debug> Debug for ImageFeatureSelector<State>

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

Formats the value using the given formatter.

impl Default for ImageFeatureSelector<Untrained>

fn default() -> Self

Returns the “default value” for a type.

impl Estimator for ImageFeatureSelector<Untrained>

type Config = ()

Configuration type for the estimator

type Error = SklearsError

Error type for the estimator

type Float = f64

The numeric type used by this estimator

fn config(&self) -> &Self::Config

Get estimator configuration

fn validate_config(&self) -> Result<(), SklearsError>

Validate estimator configuration with detailed error context

fn check_compatibility( &self, n_samples: usize, n_features: usize, ) -> Result<(), SklearsError>

Check if estimator is compatible with given data dimensions

fn metadata(&self) -> EstimatorMetadata

Get estimator metadata

impl Fit<ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>, ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>> for ImageFeatureSelector<Untrained>

type Fitted = ImageFeatureSelector<Trained>

The fitted model type

fn fit(self, x: &Array2<Float>, y: &Array1<Float>) -> SklResult<Self::Fitted>

Fit the model to the provided data with validation

fn fit_with_validation( self, x: &X, y: &Y, _x_val: Option<&X>, _y_val: Option<&Y>, ) -> Result<(Self::Fitted, FitMetrics), SklearsError>

where Self: Sized,

Fit with custom validation and early stopping

impl SelectorMixin for ImageFeatureSelector<Trained>

fn get_support(&self) -> SklResult<Array1<bool>>

Get support mask

fn transform_features(&self, indices: &[usize]) -> SklResult<Vec<usize>>

Transform by selecting features

impl Transform<ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>> for ImageFeatureSelector<Trained>

fn transform(&self, x: &Array2<Float>) -> SklResult<Array2<Float>>

Transform the input data