USING A MEDIAN FILTER AND NEURAL NETWORKS TO REDUCE IMPULSE NOISE IN COLOUR PHOTOGRAPHS

Positive and negative impulses from decoding faults or noisy channels degrade pictures in image processing. Both defects are visible and impair picture quality. Nonlinear filters may reduce impulse noise without smearing edges, unlike linear filters. Thus, nonlinear filters in image processing are crucial. Order statistics filters are common nonlinear impulse noise filters. Median filter is the most used order statistics filter since it reduces impulsive noise and preserves edges. New median-based algorithms remove jitter and random height impulse noise in speech signals, remove positive and negative impulses simultaneously in images to avoid false alarms, and study root signals and their spectral behavior. In the typical median filter, impulses induce edge jitter if they are in the same direction from the constant neighborhood. FIRMedian Hybrid (FMH) filters reduce this jitter, although impulse noise resides between the signal's lowest and maximum values. Removing signal-dependent impulse v noise and jitter in one-dimensional signals requires a novel FMH filter method and an adaptive FMH filter technique. The method replaces only noisy samples, keeping uncorrupted samples untouched. The FMH filter output is used for detection and estimation. The method efficiently removes impulsive noise from speech and audio sources while preserving their characteristics and tone quality