Introduction to the method of image noise reduction for image noise reduction

A mean value filter using neighborhood averaging is well suited for removing granular noise from images obtained by scanning. Domain averaging strongly suppresses the noise, but also causes blurring due to averaging, which is proportional to the radius of the neighborhood.

The geometric mean filter achieves a smoothness comparable to that of the arithmetic mean filter, but less image detail is lost in the filtering process.

The harmonic mean filter is better for "salt" noise, but not for "pepper" noise. It is good for other noises like Gaussian noise.

The inverse harmonic mean filter is better suited to impulse noise, but it has the disadvantage that it is necessary to know whether the noise is dark or bright in order to choose the appropriate filter order sign, and choosing the wrong order sign can be disastrous. This method retains most of the wavelet coefficients that contain the signal, so it can better maintain the image details. Wavelet analysis for image denoising has three main steps: (1) wavelet decomposition of the image signal. (2) Threshold quantization of high-frequency coefficients after hierarchical decomposition. (3) Reconstruct the image signal using two-dimensional wavelets.