Intelligent SVD-Based Noise Level Estimation Incorporating Symbiotic Organisms Search

Abstract

A simple technique for inferring the level of Additive White Gaussian Noise (AWGN) from a still image is presented in this paper. This technique exploits the effectiveness of Singular Value Decomposition (SVD) to estimate the noise level of a noisy image. It investigates the trailing sum of its singular values which contain the noise information of an image. The noise level and two additional parameters own linear dependency with the trailing sum of singular values. The two additional parameters can be experimentally obtained from a given set of noisy images. However, it becomes less satisfied in practical noise level estiation which requires a fast response. Thus, the proposed method utilizes the Symbiotic Organisms Search (SOS) to further optimize the scaling factor, regarded as additional parameter. The extensive experiments show that the proposed method offers a promising result on estimating the noise level. In addition, the estimated noise level can be further employed for the blind image denoising task.