Automatic Detection of Noise Type and Class Using Multiple Image Metrics

Researchers consider additive and multiplicative types of noises. The classes of noises are Gaussian, Salt & Pepper, which are additive type and Speckle which is multiplicative type. The most commonly used filters are mean, median and homomorphic for removing Gaussian, Salt and pepper and speckle noises respectively. If a specific filter is applied upon other type or class of noisy image, it further degrades the image instead of cleaning noise. In this research paper, a robust automatic noise detection algorithm is designed. The method developed is able to identify the type of noise i.e. additive or multiplicative and also detects the class of noise i.e. Gaussian, salt and pepper or speckle. The method compares the combinations of Kurtosis, impulse noise densities and correlation coefficient of the noisy image with predefined (experimentally found) thresholds for deciding the type and class of noise, within the noise range defined.