Lena Image Corrupted with Gaussian Noise

This document discusses the processing of a Lena image corrupted with Gaussian noise using MATLAB software. To eliminate noise, the author employs multiple techniques including T3, T5, Gaussian filtering, and median filtering. Each technique presents distinct advantages and limitations - T3 and T5 filters effectively remove Gaussian noise but may introduce some degree of blurring when eliminating image details. In contrast, Gaussian and median filters are more suitable for removing salt-and-pepper noise in images, though they demonstrate inferior performance compared to T3 and T5 in handling Gaussian noise. From an implementation perspective, the T3 and T5 filters typically involve threshold-based nonlinear filtering algorithms where pixel values are compared against neighboring values within a 3×3 or 5×5 window respectively. Gaussian filtering implements a weighted averaging process using a Gaussian kernel convolution, which can be implemented in MATLAB using the fspecial('gaussian') function followed by imfilter(). Median filtering employs the medfilt2() function to replace each pixel with the median value of its neighborhood, particularly effective for impulse noise removal. During image processing, the author also considers computational aspects such as processing time and computational cost across different filters to determine the most suitable approach. The implementation typically involves loading the original image using imread(), adding Gaussian noise with imnoise() function specifying 'gaussian' parameters, then applying respective filtering operations while comparing execution times using tic-toc functions. In summary, this paper introduces comprehensive methodologies for image processing using multiple filtering techniques, providing valuable information for readers regarding both algorithmic performance and practical MATLAB implementation considerations.