Image denoising

Image noise refers to unnecessary or redundant interference information existing in image data. In the process of digitization and transmission, digital images are often affected by the noise interference of imaging equipment and external environment. The process of reducing noise in digital images is called image denoising. Image denoising mainly consists of mean filtering, box filtering, median filtering and other methods.

Noise can be classified into several types according to different classification criteria. If the noise is divided according to its composition, it can be divided into additive noise, multiplicative noise and quantization noise.

Classification of methods for image denoising

Deep learning basedmethods:

• Convolutional Neural Network(CNN): a simple but effective denoising convolutional neural network (DnCNN, FFDnet, CBDnet) is proposed by superimposing convolutional, batch normalization and corrected linear unit (ReLU) layers .
• Non-local operations: non-local operations are incorporated into recurrent neural networks to improve denoising.
• Feature Attention: a single-stage denoising network with feature attention is proposed.

Traditional methods:

• Spatial domain methods: including mean filtering, median filtering, Gaussian filtering and bilateral filtering.
• Frequency domain methods: transform the image to the frequency domain, design filters for noise filtering, and then invert the transformation back to the spatial domain.
• Sparse representationand low-rank minimization: denoising using sparsity and low-rank properties of the image.

Optical methods:

• Diffractive visual processor: directly realizes image denoising in the process of light propagation by means of a specially designed diffraction layer, without the need for complex iterations and high-energy computation.

Your task

What kinds of noise are common in digital image processing?

Share your thoughts and ideas in the comments below.