Struct NormalizeImage 

pub struct NormalizeImage {
    pub alpha: Vec<f32>,
    pub beta: Vec<f32>,
    pub order: TensorLayout,
    pub color_order: ColorOrder,
}

Normalizes images for OCR processing.

This struct encapsulates the parameters needed to normalize images, including scaling factors, mean values, standard deviations, and channel ordering. It provides methods to apply normalization to single images or batches of images.

Fields

alpha: Vec<f32>

Scaling factors for each channel (alpha = scale / std)

beta: Vec<f32>

Offset values for each channel (beta = -mean / std)

order: TensorLayout

Tensor data layout (CHW or HWC)

color_order: ColorOrder

Color channel order (RGB or BGR)

Implementations

impl NormalizeImage

pub fn new( scale: Option<f32>, mean: Option<Vec<f32>>, std: Option<Vec<f32>>, order: Option<TensorLayout>, color_order: Option<ColorOrder>, ) -> Result<NormalizeImage, OCRError>

Creates a new NormalizeImage instance with the specified parameters.

Arguments

• scale - Optional scaling factor (defaults to 1.0/255.0)
• mean - Optional mean values for each channel (defaults to [0.485, 0.456, 0.406])
• std - Optional standard deviation values for each channel (defaults to [0.229, 0.224, 0.225])
• order - Optional tensor data layout (defaults to CHW)
• color_order - Optional color channel order (defaults to BGR)

Returns

A Result containing the new NormalizeImage instance or an OCRError if validation fails.

Errors

Returns an error if:

• Scale is less than or equal to 0
• Mean or std vectors don’t have exactly 3 elements
• Any standard deviation value is less than or equal to 0

pub fn with_color_order( scale: Option<f32>, mean: Option<Vec<f32>>, std: Option<Vec<f32>>, order: Option<TensorLayout>, color_order: Option<ColorOrder>, ) -> Result<NormalizeImage, OCRError>

Creates a new NormalizeImage instance with the specified parameters including color order.

Arguments

• scale - Optional scaling factor (defaults to 1.0/255.0)
• mean - Optional mean values for each channel (defaults to [0.485, 0.456, 0.406])
• std - Optional standard deviation values for each channel (defaults to [0.229, 0.224, 0.225])
• order - Optional tensor data layout (defaults to CHW)
• color_order - Optional color channel order (defaults to RGB)

Mean/Std Semantics

mean and std must be provided in the output channel order specified by color_order. For example, if color_order is BGR, pass mean/std as [B_mean, G_mean, R_mean].

Note: This function does not validate that mean/std values match the specified color_order. Ensuring consistency is the caller’s responsibility. If you have stats expressed in RGB order but need BGR output, prefer using NormalizeImage::with_color_order_from_rgb_stats or NormalizeImage::imagenet_bgr_from_rgb_stats which handle the reordering automatically.

Returns

A Result containing the new NormalizeImage instance or an OCRError if validation fails.

pub fn validate_config(&self) -> Result<(), OCRError>

Validates the configuration of the NormalizeImage instance.

Returns

A Result indicating success or an OCRError if validation fails.

Errors

Returns an error if:

• Alpha or beta vectors don’t have exactly 3 elements
• Any alpha or beta value is not finite

pub fn for_ocr_recognition() -> Result<NormalizeImage, OCRError>

Creates a NormalizeImage instance with parameters suitable for OCR recognition.

This creates a normalization configuration with:

• Scale: 2.0/255.0
• Mean: [1.0, 1.0, 1.0]
• Std: [1.0, 1.0, 1.0]
• Order: CHW

Returns

A Result containing the new NormalizeImage instance or an OCRError.

pub fn imagenet_rgb() -> Result<NormalizeImage, OCRError>

Creates an ImageNet-style RGB normalizer (mean/std in RGB order).

pub fn imagenet_bgr_from_rgb_stats() -> Result<NormalizeImage, OCRError>

Creates an ImageNet-style BGR normalizer from RGB stats.

This is useful for PaddlePaddle-exported models that expect BGR input, while configuration commonly provides ImageNet mean/std in RGB order.

pub fn with_color_order_from_rgb_stats( scale: Option<f32>, mean_rgb: Vec<f32>, std_rgb: Vec<f32>, order: Option<TensorLayout>, output_color_order: ColorOrder, ) -> Result<NormalizeImage, OCRError>

Builds a normalizer for a given output color_order using RGB mean/std stats.

Invariant: mean/std passed to with_color_order are interpreted in the output channel order (ColorOrder). This helper makes the conversion explicit at call sites.

pub fn apply(&self, imgs: Vec<DynamicImage>) -> Vec<Vec<f32>>

Applies normalization to a vector of images.

Arguments

• imgs - A vector of DynamicImage instances to normalize

Returns

A vector of normalized images represented as vectors of f32 values

pub fn apply_to_batch( &self, imgs: Vec<DynamicImage>, batch_tensor: &mut [f32], shapes: &[(usize, usize, usize)], ) -> Result<(), OCRError>

Applies normalization to a batch of images and stores the result in a pre-allocated tensor.

Arguments

• imgs - A vector of DynamicImage instances to normalize
• batch_tensor - A mutable slice where the normalized batch will be stored
• shapes - Shapes of the images as (channels, height, width) tuples

Returns

A Result indicating success or an OCRError if validation fails.

pub fn normalize_streaming_to_batch( &self, imgs: Vec<DynamicImage>, batch_tensor: &mut [f32], shapes: &[(usize, usize, usize)], ) -> Result<(), OCRError>

Applies normalization to a batch of images and stores the result in a pre-allocated tensor, processing images in a streaming fashion.

Arguments

• imgs - A vector of DynamicImage instances to normalize
• batch_tensor - A mutable slice where the normalized batch will be stored
• shapes - Shapes of the images as (channels, height, width) tuples

Returns

A Result indicating success or an OCRError if validation fails.

pub fn normalize_to( &self, img: DynamicImage, ) -> Result<ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, OCRError>

Normalizes a single image and returns it as a 4D tensor.

Arguments

• img - The DynamicImage to normalize

Returns

A Result containing the normalized image as a 4D tensor or an OCRError.

pub fn normalize_batch_to( &self, imgs: Vec<DynamicImage>, ) -> Result<ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, OCRError>

Normalizes a batch of images and returns them as a 4D tensor.

Arguments

• imgs - A vector of DynamicImage instances to normalize

Returns

A Result containing the normalized batch as a 4D tensor or an OCRError.

Errors

Returns an error if:

• Images in the batch don’t all have the same dimensions

Trait Implementations

impl Debug for NormalizeImage

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

Formats the value using the given formatter.