NoiseRemoval<T>

Models & Types · AiDotNet.Preprocessing.Document

NoiseRemoval - Document image noise removal with multiple algorithms.

Implements: IDisposable

For Beginners

Scanned documents often have noise (random spots, grain). This tool removes noise while keeping text clear:

  • Median: Best for salt-and-pepper noise
  • Gaussian: General smoothing
  • Bilateral: Edge-preserving smoothing
  • Morphological: Removes small artifacts

Key features:

  • Multiple noise removal algorithms
  • Edge-preserving options
  • Configurable filter sizes
  • Works with binary and grayscale images

Example usage:

How It Works

NoiseRemoval provides various filtering techniques to clean up document images by reducing noise while preserving important features like text edges.

Constructors

Constructor Summary
NoiseRemoval() Creates a new NoiseRemoval instance.

Methods

Method Summary
Tensor<T> BilateralFilter(Tensor<T> image, int kernelSize, double sigmaSpace, double sigmaColor) Applies bilateral filtering for edge-preserving smoothing.
Tensor<T> Dilate(Tensor<T> image, int kernelSize) Applies morphological dilation.
Tensor<T> Erode(Tensor<T> image, int kernelSize) Applies morphological erosion.
Tensor<T> GaussianBlur(Tensor<T> image, int kernelSize, double sigma) Applies Gaussian blur for general noise reduction.
Tensor<T> MedianFilter(Tensor<T> image, int kernelSize) Applies median filtering for salt-and-pepper noise removal.
Tensor<T> MorphologicalClosing(Tensor<T> image, int kernelSize) Applies morphological closing (dilation followed by erosion) for hole filling.
Tensor<T> MorphologicalOpening(Tensor<T> image, int kernelSize) Applies morphological opening (erosion followed by dilation) for artifact removal.
Tensor<T> Process(Tensor<T> image, NoiseRemovalMethod method) Applies noise removal to a document image.
void Dispose()