Image Denoising Using Bilateral Filter with Spea2 Optimized Parameters

Abstract

The main objective of this paper is Image Denoising while preserving edges. Denoising is often a necessary and the first step to be taken before the images data is being analyzed. Unfortunately, the images captured especially from the satellites are affected by noise which reduces the image quality. There are several reasons for this noise. They are heat generated electrons, wrong ISO settings, bad sensors, vibrations and clouds. The main aim of image Denoising is to achieve both noise reduction as well as the feature preservation. To achieve this objective, bilateral filter is being used which is a non-linear, edge-preserving and noise-reducing smoothing filter especially for images. It has two parameters namely; domain sigma and range sigma which is user defined having a predefined range[7].There are several various algorithms to achieve this image Denoising and image enhancement. In this paper, Improved Strength Pareto Evolutionary Algorithm (SPEA2) is implemented which has shown very good performance over other multi objective evolutionary algorithms like Strength Pareto Evolutionary Algorithm (SPEA) and Evolutionary Multi objective Optimization (EMO).In contrast to SPEA2, EMO does not have the features like fitness assignment, the density estimation and the archive truncation process. It also differs with SPEA as there is no single multi objective solution and has a multiple set of solutions called as Pareto efficient. SPEA2 has four Objective function named as Mean Square Error (MSE), Entropy, Structural Similarity (SSIM) and Second Derivative. In this paper, these four objective functions are used which can be defined as; Mean Square Error(MSE) is regarded as a measure of signal quality. It can be defined as the square of the difference between the noise free image and the de- noised image. Entropy is a statistical measure of randomness or disorder of the image. It is the difference between the noise free image and the Denoised image. Structural Similarity(SSIM) is used for measuring the similarity between two images. It measures the similarities of the structures between the noise free image and the Denoised image. Second Derivative can be used to detect the edges in the image. It is the convolution of the Laplacian mask and the Denoised image. This algorithm with the objective functions and the bilateral filter are implemented in MATLAB and the results of the work are represented graphically.

Keywords: Domain sigma, Range sigma, SPEA2, SSIM, MSE, Entropy, Second Derivative, Laplacian mask.