Font Size: 
Image Denoising using Enhanced Multivariance Products Representation (EMPR)
sena kaçar, Burcu Tunga

Last modified: 2020-01-15


Image denoising is a process to remove unwanted noises from the original images. Noise can occur in numerous types during storage, transmissions or acquisition of the digital image. The most popular ones are impulse, Gaussian, Speckle etc. The purpose of denoising process is to obtain high quality images from corrupted images. Researchers have improved many techniques to remove these noises in recent years. We have also tried to develop a new technique using Enhanced Multivariate Products Representation (EMPR) to increase the image quality. EMPR is a divide-and-conquer method used for decomposing multivariate data sets in terms of less variate terms with the help of support vectors for discrete structures. A digital image can be considered as a multivariate data set and thanks to EMPR algorithm, it can be decomposed into constant, univariate and bivariate components. In this study, we propose a new method using constant and univariate EMPR terms to remove Gaussian and Speckle noise from both grayscale and color images. To test the performance of the proposed method, the obtained results were compared with the well known methods in literature such as mean and median filter using Peak signal-to-noise ratio (PSNR) and Structural Similarity index (SSIM) quality measurers.