An Effective Filtering Technique for Image Denoising Using Probabilistic Principal Component Analysis (PPCA)

An image is always subjected to noise corruption at the time of capturing and transmission. In this faster ultramodern age, visual data expressed as images serve as wonderful means of communication. Yet, it is an ill-fate that the image is usually transformed after transmission due to the interruption of noise. Therefore, noise removal has turned out to be a vital and widespread challenge in image processing. Pre-processing of received image is of major consideration, prior to using it beneficially in applications. Image denoising does the manipulation of image data to render finer quality visual image to the user. Hence, a method that can eradicate salt and pepper noise, Gaussian noise and Impulse noise from the image is being proposed. The denoising technique proposed here consists of two modules to simultaneously eliminate the noise from the images in an effective way. The first module employs pixel surge model (PSM) with probabilistic principal component analysis (PPCA) for image denoising. The second module attempts to enhance the image quality by applying filters like morphological filter and region props on the results of probabilistic principal component analysis. The overall result of denoising that is produced with the proposed method is compared against the previously known method for illustrating the system effectiveness.