FUZZY IMPULSE NOISE REMOVAL IN VIDEO SEQUENCES USING BACTERIAL FORAGING ALGORITHM

Abstract

This thesis presents a new optimized fuzzy filter for denoising of color video sequences corrupted with random impulse noise. Optimization is done using Bacteria Foraging Algorithm (BFA). RGB color model is used to represent the frames of the color video. Two successive steps of filtering are used to remove the noise from the video. The first step detects the noisy pixels along with the amount of noise present and are then corrected using the median of the noise-free pixels. The output of the first step acts as an input to the second filtering step and further refines the result to give the final output. The filter presented in this paper is a 3D spatio-temporal filter that considers spatial, temporal as well as the color information. A pixel in one color component is compared with its neighboring pixels within the same frame, with the corresponding pixels in neighboring frames and also with the pixels in other two color components. Six fuzzy membership functions are defined which are then applied to a fuzzy rule. The fuzzy rule decides whether the given pixel is noisy or noisefree. Mean Square Error (MSE) is used as an objective function, which is optimized using the bacterial foraging algorithm to learn the parameters of membership functions. Finally, the correction term for each color component of each frame is calculated which is added to the original noisy frame component to remove the noise. Only noisy pixels are corrected while the noisefree pixels are left untouched. The experimental result on several color video sequences proves that the impulse noise is efficiently removed by the proposed fuzzy filter.