Gaussian and Impulse Noise Identification from Image Using Frequency Domain Analysis

Classifying the various types of noise and reconstructing de-noisy images is one of the challenging issues in the field of image processing. Noise, compression artifacts, blurring, and sensor limits are a few examples of the many causes of distortion. To increase data quality and signal processing in a variety of fields, such as pictures, audio processing, and telecommunications, precise noise type identification is necessary for the development of noise reduction algorithms suited to particular applications. The statistical characteristics of Gaussian noise have a Gaussian (normal) distribution. It is usually enhanced to the overall variability of a signal and frequently linked to environmental factors. On the other hand, abrupt, high-intensity spikes or signal disruptions are characteristic of impulse noise. In this paper, the frequency domain analysis approach is utilized to find Gaussian blur and impulse noise in an image. Signals or time-domain data can be converted into its frequency domain representation using a computing process called the Fourier transform (FFT). This study suggests a two-step method for classifying different types of noise. First, the frequency spectrum is shown by applying the FFT to the noisy input. Then, statistical measures such as mean, standard deviation, and z-score of spectrum are computed. FFT is used to characterize the spectral characteristics of Gaussian and impulsive noise from different images.