ELIMINATION OF IMPULSIVE NOISE IN IMAGES BY MEANS OF THE USE OF SUPPORT VECTOR MACHINES
Hilario Gómez-Moreno, Saturnino Maldonado Bascón, Manuel Utrilla Manso, Pilar Martín Martín · 2001
In this work we present an efficient way to cancel the impulsive noise into images by using the Support Vector Machines (SVM). The suppression of impulsive noise is a classic problem in nonlinear processing, and the SVM are especially useful in this type of processing. In this new approach we use the classification and the regression based on SVM. By using the classifier we select the noisy pixels into the images and by using the regression we obtain a reconstruction value based on the neighboring pixels. The results obtained are comparable and, a lot of times, better than those from another "state-of-art " techniques. Besides, this new technique can be applied successfully to images with high noise ratios while maintaining the visual quality and the low reconstruction error. 1.